Playerbots rewrite proposal

Create the first tracer through a reproducible pinned Module and Package fixture. Drive attack, interaction, movement, and cast-start through real core operations, return typed outcomes to the Package, and capture the blocked-target and out-of-range-heal failures that the rewrite must fix.

Acceptance criteria

• The fixture records pinned core and Package commits plus the actual content and geometry input revisions, labels seeded and client-derived provenance separately, records unavailable inputs, and reproduces the same initial state and outcome.
• The fixture loads and runs the real Module with the real Package; it does not substitute a simulated gameplay engine.
• Attack acceptance is distinct from damage and quest progress, and an attempted attack with no hit cannot clear a quest progress clock.
• Movement reports complete, safe partial, blocked, or explicit unknown-coverage evidence with route or waypoint advancement where applicable.
• Interaction and cast-start return typed success, waiting, Refusal, cancellation, and expiry outcomes supported by their owning operations.
• A long bot cast uses the normal core cast lifecycle and completes once without restart on later decision ticks.
• The blocked target and poor-range heal cases have reproducible inputs, observed outcomes, and enough evidence to compare later runner behavior.
• A written extraction comparison chooses the smaller total implementation and records why the full AzerothCore runtime is not used.

Verified criteria: 8/8.

The fixture records pinned core and Package commits plus the actual content and geometry input revisions, labels seeded and client-derived provenance separately, records unavailable inputs, and reproduces the same initial state and outcome.

Reviewed by Codex orchestrator with independent GPT-6 Astra Spec review.

• All 16 final CI input records name clean core fe9caef6eb0d132c45a5d151827ab6fe64925f79 and GitHub tested Package merge 744732b50cc8135d3a06e598e07e26aa8633a434, whose parents are main c32b2fb and reviewed Package c97464f6be211525a3b36a2ee3824cf259edbdd1. Content is playerbots-seeded-actions-v1; geometry is playerbots-synthetic-nav-v1. Imported content and client geometry are explicitly unavailable.
• CI Module BLAKE3: a4e921d874c49ada97f9893b4c9b41847ffdb994c670dcd3ea5673f0fab466fd. Package BLAKE3: 1b4de3cdafc56407264c216a3accda917117277914b2deffb4fbd0c3efaf29f6. The local focused Module has a different installed inventory; byte equality across those builds is not claimed.
• Input manifest: https://github.com/LyraCoreProject/packages/blob/c97464f6be211525a3b36a2ee3824cf259edbdd1/playerbots/fixtures/actions.json
• CI input and outcome artifacts: https://github.com/LyraCoreProject/packages/actions/runs/34229152446
• The final run against merged core be3fa67d also records 16 clean input and outcome pairs. Tested Package merge b451105562bbb5deab45ceff9919b9e002aa240d has parents c32b2fb and reviewed head 0781d7e5. Module and Package identities match the preceding CI run. Artifacts: https://github.com/LyraCoreProject/packages/actions/runs/34230810246

The fixture loads and runs the real Module with the real Package; it does not substitute a simulated gameplay engine.

Reviewed by Codex orchestrator with independent GPT-6 Astra Spec review.

• The local fixture passed all 16 cases through compiled Module Wasm, the real Package, and private SpacetimeDB 2.7.1 Standalone nodes. The final CI run below also includes the later repeated-setup checks.
• The same behavior passed in Package CI: https://github.com/LyraCoreProject/packages/actions/runs/34226546164/job/102061988209
• Final fixture changes also passed all 16 cases in 166.26 seconds: https://github.com/LyraCoreProject/packages/actions/runs/34229152446/job/102070649535
• The same 16 fixtures passed against merged core main before the Package merge: https://github.com/LyraCoreProject/packages/actions/runs/34230810246/job/102076215519

Attack acceptance is distinct from damage and quest progress, and an attempted attack with no hit cannot clear a quest progress clock.

Reviewed by Codex orchestrator with independent GPT-6 Astra Spec review.

• playerbots_an_accepted_attack_without_a_hit_does_not_clear_quest_stall observes accepted engagement, full target health, zero quest credit, and an unchanged stall clock behind a synthetic wall. Its positive phase uses the real lethal-hit pipeline to award credit; it does not claim that the bot hit through the wall.
• https://github.com/LyraCoreProject/LyraCore/blob/fe9caef6eb0d132c45a5d151827ab6fe64925f79/module/tests/playerbots_rewrite.rs

Movement reports complete, safe partial, blocked, or explicit unknown-coverage evidence with route or waypoint advancement where applicable.

Reviewed by Codex orchestrator with independent GPT-6 Astra Spec review.

• The complete-route and partial-route cases distinguish a proposed leg from observed arrival, retain a waypoint and search-work count, and record Unknown coverage. The blocked-target case holds position. Shared navigation tests passed 34 cases; VerifiedCells only attests consulted derived rows.
• https://github.com/LyraCoreProject/LyraCore/blob/fe9caef6eb0d132c45a5d151827ab6fe64925f79/module/tests/playerbots_rewrite.rs

Interaction and cast-start return typed success, waiting, Refusal, cancellation, and expiry outcomes supported by their owning operations.

Reviewed by Codex orchestrator with independent GPT-6 Astra Spec review.

• The fixture covers typed cast start/wait/resolve/refusal/cancel/expire outcomes, synchronous quest completion/refusal, full-inventory and invalid-choice atomicity, source-item refusal, and dead-attacker refusal. All 16 final cases passed.
• https://github.com/LyraCoreProject/LyraCore/blob/fe9caef6eb0d132c45a5d151827ab6fe64925f79/module/tests/playerbots_rewrite.rs

A long bot cast uses the normal core cast lifecycle and completes once without restart on later decision ticks.

Reviewed by Codex orchestrator with independent GPT-6 Astra Spec review.

• playerbots_a_long_bot_cast_waits_for_core_completion retains one five-second cast through repeated requests and normal decision ticks, observes one START and one GO, and verifies the heal. The owning begin operation returns its exact result directly.
• https://github.com/LyraCoreProject/LyraCore/blob/fe9caef6eb0d132c45a5d151827ab6fe64925f79/module/tests/playerbots_rewrite.rs

The blocked target and poor-range heal cases have reproducible inputs, observed outcomes, and enough evidence to compare later runner behavior.

Reviewed by Codex orchestrator with independent GPT-6 Astra Spec review.

• The blocked-target scenario and playerbots_an_out_of_range_heal_reports_the_owning_cast_gate record reproducible inputs and typed observations. The poor-range target starts at x=1400 while the bot starts at x=1200. Actions retain target, route, clipping, coverage, cast identity, and Refusal.
• https://github.com/LyraCoreProject/LyraCore/blob/fe9caef6eb0d132c45a5d151827ab6fe64925f79/module/tests/playerbots_rewrite.rs

A written extraction comparison chooses the smaller total implementation and records why the full AzerothCore runtime is not used.

Reviewed by Codex orchestrator with independent GPT-6 Astra Spec review.

• The extraction comparison in the Package README chooses typed Rust decisions and existing Module operations. It compares the 2,967-line selection-kernel inventory with the 243,560-line C++ runtime and records host-pointer, packet, and WotLK dependencies. https://github.com/LyraCoreProject/packages/blob/c97464f6be211525a3b36a2ee3824cf259edbdd1/playerbots/README.md
• https://github.com/LyraCoreProject/LyraCore/pull/482#issuecomment-5585257477

Execution model: gpt-6-astra, high effort. Repositories: LyraCore Module action and navigation contracts, Official Package Collection playerbots fixture, Local durable integration tests.

Recorded evidence

• Baseline on core 95e6ad8 and Package c32b2fb: both ignored personality durable tests passed in a private loopback Standalone. This proves the existing Package fixture, not the new outcome contract.
• Final local union passed 1,621 Module unit tests normally, 1,622 with debug reducers, and 32 durable cases. Default-core SpacetimeDB 2.7.1 generation matched all 995 committed Gateway binding files. Independent standards findings are resolved or explicitly accepted; Spec review found no confirmed defects.
• The final Package CI run passed all 16 strengthened fixtures against core fe9caef6. Repeated interaction setup now clears only its reserved target and quest rows, including stale movement and engagement. Evidence: https://github.com/LyraCoreProject/packages/actions/runs/34229152446/job/102070649535
• Core #482 merged as be3fa67d0f0c24749230560544a3e8e8b577f61d after all checks passed, including the full durable job: https://github.com/LyraCoreProject/LyraCore/actions/runs/34229150494/job/102070610422
• Final Package checks against merged core main passed, including all 16 durable fixtures in 167.90 seconds. Package #25 merged as 5724de1660a2628e88320914bdd5abb0c69da517. CI: https://github.com/LyraCoreProject/packages/actions/runs/34230810246/job/102076215519

Current notes

• All eight source acceptance criteria have reviewed evidence. Core #482 and Package #25 are merged.
• Package CI was restored to core main after the core merge. All final checks passed before the Package merge.
• Both full CodeRabbit reviews completed. The repeated-setup finding is fixed; the core CI coverage question is resolved by the paired Package durable job. Final incremental automated review was rate limited. Independent review covered the final fixture changes. All core checks passed.
• The flat observation row remains intentional for SQL diagnosis. Action-specific request functions and a private recorder own production writes; PB-002 owns typed runner candidates and foreground control.
• The nullable quest-credit initialization fixture preserves an existing stall. A full older-Wasm upgrade was not claimed. The new Package schema remains subject to the repository human review requirement before shipping.

Build the shared decision runner around typed facts and outcomes. Retain one objective, arbitrate one foreground action, schedule bots in stable bounded batches, and explain each meaningful choice through the Package's normal entry point.

Acceptance criteria

• Fixed facts and explicit time produce the same target-bearing candidate order and chosen action.
• Strategies compose supported triggers, defaults, and priority rules without a global string registry or a second authoritative state machine.
• Prerequisites, alternatives, and continuers have explicit depth, transition, candidate, and route-work limits with cycle refusal.
• Only one foreground movement or cast action owns the bot, and a higher-priority action explicitly cancels incompatible work.
• The durable objective records identity, destination, stage, deadline, last verified progress, bounded failures, deferred destinations, and catalog revision.
• Separate movement, combat, cast, and quest progress evidence prevents an accepted action from masquerading as completion.
• Due bots run in stable indexed order under a per-pass limit; excess work remains due, gets a fair next turn, and records scheduler lag.
• The explanation read reports the current objective and target, chosen action and reason, last outcome, progress age, retry count, next eligible time, and relevant revision with bounded history.
• The additive state migration defines defaults and backfill and adds a durable controller selector before any bot can use the new runner. Record-only mode emits no gameplay actions; cohort mode has one eligible controller and supports freeze.

Verified criteria: 9/9.

Fixed facts and explicit time produce the same target-bearing candidate order and chosen action.

Reviewed by Codex orchestrator with independent GPT-6 Astra Spec and GPT-5.6 Sol Standards reviews.

• The pure decision tests permute input order, rank full target-bearing identities, and retain only the highest-priority offer for each identity. All ten decision cases passed in the final library runs.
• https://github.com/LyraCoreProject/packages/blob/62c459770eb5b2209b7cfbf38e54857454fa69d2/playerbots/src/decision.rs

Strategies compose supported triggers, defaults, and priority rules without a global string registry or a second authoritative state machine.

Reviewed by Codex orchestrator with independent GPT-6 Astra Spec and GPT-5.6 Sol Standards reviews.

• Typed Strategy tests cover triggers, defaults, and integer priority adjustment. The common runner stamps its durable objective identity into root and nested candidates before selection.
• https://github.com/LyraCoreProject/packages/blob/62c459770eb5b2209b7cfbf38e54857454fa69d2/playerbots/src/decision.rs
• https://github.com/LyraCoreProject/packages/blob/62c459770eb5b2209b7cfbf38e54857454fa69d2/playerbots/src/runner.rs

Prerequisites, alternatives, and continuers have explicit depth, transition, candidate, and route-work limits with cycle refusal.

Reviewed by Codex orchestrator with independent GPT-6 Astra Spec and GPT-5.6 Sol Standards reviews.

• The shared limits allow 24 candidates, depth 4, 16 transitions, and one route reservation using the core expansion limit. Separate tests cover each limit, cycles, and repeated cycle refusals inside completed prerequisites.
• https://github.com/LyraCoreProject/packages/blob/62c459770eb5b2209b7cfbf38e54857454fa69d2/playerbots/src/decision.rs
• Recovery facts use a paged index and exact learned-spell membership. Runtime regressions find a usable heal beyond 25 unlearned rotations and 300 other learned spells without an unbounded eligibility scan.

Only one foreground movement or cast action owns the bot, and a higher-priority action explicitly cancels incompatible work.

Reviewed by Codex orchestrator with independent GPT-6 Astra Spec and GPT-5.6 Sol Standards reviews.

• The final real runner suite proves cast retention through pushback, explicit survival preemption, changed-destination cancellation, and freeze. The Account ownership case preserves a later human cast while the bot relinquishes control. All 21 runner cases passed.
• https://github.com/LyraCoreProject/LyraCore/blob/4f450074d76f8982d1a7d61898a00cf632132737/module/tests/playerbots_rewrite/runner.rs

The durable objective records identity, destination, stage, deadline, last verified progress, bounded failures, deferred destinations, and catalog revision.

Reviewed by Codex orchestrator with independent GPT-6 Astra Spec and GPT-5.6 Sol Standards reviews.

• The retained ReturnHome objective records destination, identity, stage, deadline, last verified progress, and catalog revision. Blocked travel defers its destination and caps failures and deferrals at four entries. Real tests cover actual arrival, deferred travel under low health, and deadline expiry.
• https://github.com/LyraCoreProject/packages/blob/62c459770eb5b2209b7cfbf38e54857454fa69d2/playerbots/src/runner.rs
• https://github.com/LyraCoreProject/LyraCore/blob/4f450074d76f8982d1a7d61898a00cf632132737/module/tests/playerbots_rewrite/runner.rs

Separate movement, combat, cast, and quest progress evidence prevents an accepted action from masquerading as completion.

Reviewed by Codex orchestrator with independent GPT-6 Astra Spec and GPT-5.6 Sol Standards reviews.

• The final tactical movement case moves away from home for more than ten seconds, records movement progress, and leaves the home progress clock unchanged. Defense acceptance does not count as damage. Cast completion uses the exact core scheduled identity. PB-001 quest-credit and no-hit cases still pass.
• https://github.com/LyraCoreProject/LyraCore/blob/4f450074d76f8982d1a7d61898a00cf632132737/module/tests/playerbots_rewrite/runner.rs
• https://github.com/LyraCoreProject/LyraCore/blob/4f450074d76f8982d1a7d61898a00cf632132737/module/tests/playerbots_rewrite.rs

Due bots run in stable indexed order under a per-pass limit; excess work remains due, gets a fair next turn, and records scheduler lag.

Reviewed by Codex orchestrator with independent GPT-6 Astra Spec and GPT-5.6 Sol Standards reviews.

• The real 25-bot case proves stable indexed processing in batches of 16 then 9. Excess work remains due and records lag. The populated migration also verifies bounded lazy backfill.
• https://github.com/LyraCoreProject/LyraCore/blob/4f450074d76f8982d1a7d61898a00cf632132737/module/tests/playerbots_rewrite/runner.rs

The explanation read reports the current objective and target, chosen action and reason, last outcome, progress age, retry count, next eligible time, and relevant revision with bounded history.

Reviewed by Codex orchestrator with independent GPT-6 Astra Spec and GPT-5.6 Sol Standards reviews.

• The public runner row exposes the objective, target-bearing chosen action and reason, last outcome, progress age, retry count, next eligible time, and revision. Its real lifetime case caps history at eight entries and deletes Character-owned state.
• https://github.com/LyraCoreProject/packages/blob/62c459770eb5b2209b7cfbf38e54857454fa69d2/playerbots/src/runner.rs
• https://github.com/LyraCoreProject/LyraCore/blob/4f450074d76f8982d1a7d61898a00cf632132737/module/tests/playerbots_rewrite/runner.rs
• The separate Recovery Scan row exposes its cursor, bounded rows scanned, retained rotation, current stage, and last completed result. Missing remains valid during a subsequent Pending scan.

The additive state migration defines defaults and backfill and adds a durable controller selector before any bot can use the new runner. Record-only mode emits no gameplay actions; cohort mode has one eligible controller and supports freeze.

Reviewed by Codex orchestrator with independent GPT-6 Astra Spec and GPT-5.6 Sol Standards reviews.

• The final populated upgrade publishes a preceding Wasm, creates bot roster, goals, observations, and quest rows, then upgrades the same private database. Legacy and scheduler-lag defaults survive. Normal tick and hook cases prove RecordOnly and Frozen suppression and one Cohort controller. Five real core consent cases cover current ownership, invitation admission, deletion, and Transfer.
• https://github.com/LyraCoreProject/LyraCore/blob/4f450074d76f8982d1a7d61898a00cf632132737/module/tests/playerbots_rewrite/runner.rs
• https://github.com/LyraCoreProject/LyraCore/blob/4f450074d76f8982d1a7d61898a00cf632132737/module/tests/sessionless_consent.rs

Execution model: gpt-6-astra, high effort. Repositories: Official Package Collection decision and durable state, LyraCore Module scheduling and diagnostics, Pure decision and durable integration tests.

Recorded evidence

• All nine criteria have reviewed source and runtime evidence. Final reviewed heads are core 4f450074d76f8982d1a7d61898a00cf632132737 and Package 62c459770eb5b2209b7cfbf38e54857454fa69d2. The complete local correction run used clean core a4b5518 and Package dba96bb; the later changes strengthen one test and restore CI references.
• The full Package union passed 1,631 library tests, 1,632 with debug reducers, all 37 action and runner cases, two personality cases, and five core consent cases. This includes a populated preceding-Wasm migration.
• All 37 runtime input records agree on Module BLAKE3 dba4d702a0320e360506f9698fa52e0ef7210ce1e8f9573928599a8e3e032279 and Package Content Identity 33fcb8a217aad84f46f9ad65efdad23be24bad8e29828777de7c7e6469417161. Saved Wasm: 7,590,466 bytes; SHA256 c44a91ee92bb22ebce3de07e39caf9d7028794460a855c3b1d1f9329e473a2ae.
• Final Core #486 CI passed after the test-only update, including the full durable suite and all 1,000 default Gateway binding files. The exact spellbook index needs no generated binding change. CI: https://github.com/LyraCoreProject/LyraCore/actions/runs/34248547507/job/102136730760
• Corrected paired Package CI passed all 37 cases against core a4b5518. Package branch dba96bb ran as synthetic merge e786ee50bb06fc1902b9f5a8e9514669f2d0689e. All inputs are clean and the Package content identity matches local. CI Wasm BLAKE3 is 80efd76cef0d52ee0342f9bc1fe1ee011bcbace0eac7e67e68c944cd427bf659. CI: https://github.com/LyraCoreProject/packages/actions/runs/34245772629/job/102127317130
• The populated local upgrade uses preceding core be3fa67d0f0c24749230560544a3e8e8b577f61d and Package 5724de1660a2628e88320914bdd5abb0c69da517. Its Wasm BLAKE3 is 1662b0963dc0e03c29ddde985f08c0ec5a6f392e899a80eac9d7789a85b1b5a9. Preceding local inventory has example, fire_nova, and playerbots; current inventory also has dungeons. CI independently builds the full preceding collection and records its own Wasm identity.
• Six added regressions cover late eligible heals, retained healing eligibility, health-ratio overflow, isolated fixture timing, navigation grid edges, and a completed missing heal result that permits home movement during rescans. Independent Spec and Standards reviews cleared the final source; the orchestrator reviewed the workflow-only pin update.
• The final core change, 4f450074d76f8982d1a7d61898a00cf632132737, strengthens the grid cleanup test. Its focused durable run proves that each grid corner loses all four nearby chunks while a distant chunk survives. Independent Spec and Standards reviews cleared this test-only delta, and CodeRabbit marked its finding addressed. Gameplay source is unchanged from the 37-case run.
• The focused grid run records clean core 4f45007 and unchanged Package dba96bb. Its Wasm BLAKE3 is 18a320609eebb7a163154d891953decbbc5bdaa401e81d339acb06dd73eaa9e9 and SHA256 is 713ad9a2401029fd10e184c91da3a4eeee1c7e402ad206bf59b598e4afb55a0e. Package Content Identity matches the complete local suite.
• CodeRabbit completed its Package correction review at dba96bb. Its remaining concern assumed absent consent blocks Legacy bots. Core action admission does not read consent, and group admission suppresses only an existing false row. A durable case proves admission before a consent row exists. The orchestrator independently verified this and the implementer resolved the comment with immutable source and behavior evidence.
• Final PR CI passed all 37 cases against merged core e6a755db0a150bbf73ad97b972fe829f20f6816c and finalized its artifact upload. Actual tested Package merge d8d6d7c6574948c07d277ae55e1c899e7721ff95 has parents 5724de1660a2628e88320914bdd5abb0c69da517 and 62c459770eb5b2209b7cfbf38e54857454fa69d2. CI: https://github.com/LyraCoreProject/packages/actions/runs/34250561420/job/102143672288
• The orchestrator independently checked all 37 clean final runtime records. Module Wasm BLAKE3 is 80efd76cef0d52ee0342f9bc1fe1ee011bcbace0eac7e67e68c944cd427bf659. Package Content Identity is 33fcb8a217aad84f46f9ad65efdad23be24bad8e29828777de7c7e6469417161. The strengthened grid assertion and populated migration passed.
• Package #26 merged as 155c9e401afb06d5731acedf8fc35a81dbe4aaa6. Its Git tree e2558a9cf421a74f79497ef361dbbfceb911a0d1 exactly matches both the reviewed branch and the successful PR merge build.
• The separate push run also passed all 37 tests, then GitHub returned HTTP 403 while finalizing its artifact upload. One retry passed all 37 cases and finalized its artifact successfully. The orchestrator verified its 37 clean inputs at core e6a755d and Package 62c4597, with the same Wasm and Package identities as the successful PR run. This closes the service failure without a source change. Retry: https://github.com/LyraCoreProject/packages/actions/runs/34250556649/job/102147886460

Current notes

• SpacetimeDB 2.7.1 allows appended enum variants only when old variants stay compatible and storage size and alignment stay unchanged. Prove populated-state migration for the runner schema. Planner source: https://github.com/clockworklabs/SpacetimeDB/blob/v2.7.1/crates/schema/src/auto_migrate.rs#L826-L895
• Direct damage can delay a pending cast while retaining its scheduled identity. Observe its current due time before deciding that completion is late; the objective deadline is a separate policy limit.
• Record-only means no gameplay actions from either runner for the selected bot. Freeze must stop retained foreground work. Verify both through the normal Package tick entry.
• The first runner behavior retains a home destination through defense and self-heal interruptions. Companion policy and quest admission remain in their dependent tickets.
• Control must cover damage and invitation hooks as well as the tick. The Gateway also auto-accepts invitations for session-less Characters. The core-owned consent check covers that path for record-only and frozen bots.
• The consent check acknowledges current World Shard state before Realm-core group actions. An action admitted before a concurrent freeze may finish; cross-Shard atomicity is not claimed. Controller selection clears unclaimed Group Intents in the same transaction.
• The common runner entry checks current Account ownership. Relinquishment releases only the bot's observed cast or movement identity and preserves a later human action.
• Package schema changes still require human review before production shipping. Agent review and source merge do not satisfy that production gate.
• All three PB-002 implementation PRs are merged. Full source reviews and the final workflow review have no unresolved finding. Both final PR and push CI runs passed and saved their artifacts.
• Recovery Scan advances through at most 24 indexed rotations per pass and rechecks at most two retained rotations through exact spellbook lookups. Completed Missing remains conclusive while a new scan is Pending, so recovery cannot repeatedly interrupt useful movement.
• The missing-result regression observes ReturnHome selection and a real emitted spline while movement scheduling is disabled. Separate arrival and tactical cases prove actual position advancement.
• The Recovery Scan is disposable on Transfer and restarts against current content and learned spells. PB-010 still owns transfer of durable objectives and companion intent.

Complete the first human-led companion loop. A Priest follows its party leader, retains an injured ally as the heal target, repairs range or line of sight, casts through the real lifecycle, handles interruption, and resumes following.

Acceptance criteria

• Joining a human-led party suspends autonomous travel and selects companion Strategies from durable party facts.
• The Priest follows a moving leader without starting a fight against an unrelated nearby enemy.
• An injured ally outranks ordinary following when the Priest has a usable heal and the survival policy permits it.
• Poor range or line of sight produces a movement prerequisite that retains the injured ally and finds a reachable casting position or records a specific wait.
• The heal starts one real timed cast, does not restart on repeated decisions, and observes completion or interruption through typed outcomes.
• After healing, cancellation, or a recoverable interruption, the Priest resumes following without losing party control state.
• Death and resurrection preserve the companion role and lead to regrouping rather than a permanent solo objective change.

Verified criteria: 7/7.

Joining a human-led party suspends autonomous travel and selects companion Strategies from durable party facts.

Reviewed by Independent final Spec and Standards reviewers; orchestrator verified all 57 clean input records.

• priest_follows_a_moving_human_leader_without_pulling and missing_group_parent_holds_the_companion_objective prove party selection and preserved control.
• Final Core main CI 34286733126 passed all 20 companion and 37 shared cases, plus 1,633 Module library tests. The orchestrator verified 57 clean case input records at Core 25c4bd9dbff61c0dab23cbac63e8e4b771a04593. Tested Package merge 2fb24b89b0950025a86bf3d13d9862fd21106da4 has the same tree as reviewed 7f23a4ba93eae5f560c63960b49622cde6042b6d and merged 284c0ab56057da75e18112a6308a0696e5ccdbaa. Both implementation PRs are merged. Independent final Spec and Standards reviews approved.

The Priest follows a moving leader without starting a fight against an unrelated nearby enemy.

Reviewed by Independent final Spec and Standards reviewers; orchestrator verified all 57 clean input records.

• priest_follows_a_moving_human_leader_without_pulling proves actual follow movement with no unrelated attack.
• Final Core main CI 34286733126 passed all 20 companion and 37 shared cases, plus 1,633 Module library tests. The orchestrator verified 57 clean case input records at Core 25c4bd9dbff61c0dab23cbac63e8e4b771a04593. Tested Package merge 2fb24b89b0950025a86bf3d13d9862fd21106da4 has the same tree as reviewed 7f23a4ba93eae5f560c63960b49622cde6042b6d and merged 284c0ab56057da75e18112a6308a0696e5ccdbaa. Both implementation PRs are merged. Independent final Spec and Standards reviews approved.

An injured ally outranks ordinary following when the Priest has a usable heal and the survival policy permits it.

Reviewed by Independent final Spec and Standards reviewers; orchestrator verified all 57 clean input records.

• priest_retains_one_ally_cast_while_the_leader_moves_then_resumes_follow, unsupported_channel_does_not_hide_a_supported_learned_heal, and low_health_at_the_reached_leader_uses_recovery_instead_of_holding prove priority and usable recovery.
• Final Core main CI 34286733126 passed all 20 companion and 37 shared cases, plus 1,633 Module library tests. The orchestrator verified 57 clean case input records at Core 25c4bd9dbff61c0dab23cbac63e8e4b771a04593. Tested Package merge 2fb24b89b0950025a86bf3d13d9862fd21106da4 has the same tree as reviewed 7f23a4ba93eae5f560c63960b49622cde6042b6d and merged 284c0ab56057da75e18112a6308a0696e5ccdbaa. Both implementation PRs are merged. Independent final Spec and Standards reviews approved.

Poor range or line of sight produces a movement prerequisite that retains the injured ally and finds a reachable casting position or records a specific wait.

Reviewed by Independent final Spec and Standards reviewers; orchestrator verified all 57 clean input records.

• poor_range_selects_the_injured_allys_casting_position, targeted_los_gate_matches_client_and_bot_casts, and casting_position_retains_one_injured_ally_across_movement_legs prove movement prerequisites and retained target identity.
• Final Core main CI 34286733126 passed all 20 companion and 37 shared cases, plus 1,633 Module library tests. The orchestrator verified 57 clean case input records at Core 25c4bd9dbff61c0dab23cbac63e8e4b771a04593. Tested Package merge 2fb24b89b0950025a86bf3d13d9862fd21106da4 has the same tree as reviewed 7f23a4ba93eae5f560c63960b49622cde6042b6d and merged 284c0ab56057da75e18112a6308a0696e5ccdbaa. Both implementation PRs are merged. Independent final Spec and Standards reviews approved.

The heal starts one real timed cast, does not restart on repeated decisions, and observes completion or interruption through typed outcomes.

Reviewed by Independent final Spec and Standards reviewers; orchestrator verified all 57 clean input records.

• priest_retains_one_ally_cast_while_the_leader_moves_then_resumes_follow proves one real timed cast through repeated decisions. Cancellation and completion-time LOS cases prove typed release paths.
• Final Core main CI 34286733126 passed all 20 companion and 37 shared cases, plus 1,633 Module library tests. The orchestrator verified 57 clean case input records at Core 25c4bd9dbff61c0dab23cbac63e8e4b771a04593. Tested Package merge 2fb24b89b0950025a86bf3d13d9862fd21106da4 has the same tree as reviewed 7f23a4ba93eae5f560c63960b49622cde6042b6d and merged 284c0ab56057da75e18112a6308a0696e5ccdbaa. Both implementation PRs are merged. Independent final Spec and Standards reviews approved.

After healing, cancellation, or a recoverable interruption, the Priest resumes following without losing party control state.

Reviewed by Independent final Spec and Standards reviewers; orchestrator verified all 57 clean input records.

• explicit_cancellation_releases_the_heal_and_resumes_follow and completion_time_los_refusal_releases_the_heal_and_resumes_follow prove recovery to party following.
• Final Core main CI 34286733126 passed all 20 companion and 37 shared cases, plus 1,633 Module library tests. The orchestrator verified 57 clean case input records at Core 25c4bd9dbff61c0dab23cbac63e8e4b771a04593. Tested Package merge 2fb24b89b0950025a86bf3d13d9862fd21106da4 has the same tree as reviewed 7f23a4ba93eae5f560c63960b49622cde6042b6d and merged 284c0ab56057da75e18112a6308a0696e5ccdbaa. Both implementation PRs are merged. Independent final Spec and Standards reviews approved.

Death and resurrection preserve the companion role and lead to regrouping rather than a permanent solo objective change.

Reviewed by Independent final Spec and Standards reviewers; orchestrator verified all 57 clean input records.

• death_and_resurrection_preserve_role_then_regroup proves role retention and regrouping. populated_pb002_runner_state_upgrades_with_objective_and_foreground_intact proves additive state migration.
• Final Core main CI 34286733126 passed all 20 companion and 37 shared cases, plus 1,633 Module library tests. The orchestrator verified 57 clean case input records at Core 25c4bd9dbff61c0dab23cbac63e8e4b771a04593. Tested Package merge 2fb24b89b0950025a86bf3d13d9862fd21106da4 has the same tree as reviewed 7f23a4ba93eae5f560c63960b49622cde6042b6d and merged 284c0ab56057da75e18112a6308a0696e5ccdbaa. Both implementation PRs are merged. Independent final Spec and Standards reviews approved.

Execution model: gpt-5.6-sol, high effort. Repositories: Official Package Collection companion support behavior, LyraCore Module cast and party fact reads, Durable companion scenarios.

Recorded evidence

• All 57 final local cases passed at clean core 8306ccb4c5951fe4c39dee69717e446ad8af0bde and Package b9d09f9c6ab0033c66c530642412f98a6dc3f57d. The companion target passed 20 cases in 144.47 seconds; the action and runner target passed 37 in 313.80 seconds. Both independent final source reviews approved.
• The orchestrator verified all 57 per-case input records against the exact clean pair. The saved Module is 7,518,238 bytes with SHA256 59ee91daf3aee158443afdc8f1db33b4288c3ba7432338523219210baab7e54b and BLAKE3 abdad11fa26c0c05e311b7f394f44069b00b27934beaf0a640bc215a8d5fd7e4. Package Content Identity is ccbc1296a70ebb3df7f53bb57e0c91b64eaf56de5c0e80ad31d919e2c3dc0fd6.
• The suite includes populated PB-002 migration, retained ally selection across observed movement legs, normal learned-spell admission, actual cast completion, cancellation, death and regrouping. Earlier failed runs remain separate evidence for their exact pins. Final PR CI and both merges remain required. Geometry is synthetic; imported-world and attended-client evidence remain later release gates.
• Final PR CI at core 8306ccb4 and Package b9d09f9 passed 37/37 shared cases and 19/20 companion cases. The movement case counted short automatic segments differently across hosts. The revised fixture controls only decision timing, observes each actual spline endpoint, and retains the ally across two to six legs. Its focused run and the full PR suite are pending at the newer heads. The migration fixture now stages a longer real cast and compares the pending handle across publish.
• The controlled-spline fixture passed its focused retention and populated migration cases at core b3bf5a7e and Package 87468ec4. Final PR CI passed all 57 cases at that pair. Independent Spec review then found an initial controller-selection scheduling gap. The revised fixture selects and parks in one transaction; its focused verification and updated CI are running.
• The final atomic-selection pair, core 624edcdfde9c607a22c9ddc751b7c40caf6c7018 and Package 4741c973910a78127bf722bde274c79556b75306, passed retention in 14.14 seconds and populated migration in 6.59 seconds. The orchestrator verified both clean input records and the saved 7,522,611-byte Module, SHA256 2e43ae814df3b4e472113736565c1bf263ce7a99cc41749660d09d37d526cf12. Both independent final delta reviews approve. Updated full CI and integration remain pending.
• Core #489 merged as 25c4bd9dbff61c0dab23cbac63e8e4b771a04593 after all checks and both final reviews passed. Its tree equals reviewed core 624edcdf. Package CI passed 20 companion and 37 shared cases in 149.60 and 347.18 seconds. All 57 input records are clean, and tested merge f9d04bccaae646bd5b2d8490b730f11f8fabe9b1 has the same tree as Package 4741c973. Package #27 now restores current Core main coverage; its final workflow CI and merge remain pending.
• Final Core main CI 34286733126 passed all 20 companion and 37 shared cases, plus 1,633 Module library tests. The orchestrator verified 57 clean case input records at Core 25c4bd9dbff61c0dab23cbac63e8e4b771a04593. Tested Package merge 2fb24b89b0950025a86bf3d13d9862fd21106da4 has the same tree as reviewed 7f23a4ba93eae5f560c63960b49622cde6042b6d and merged 284c0ab56057da75e18112a6308a0696e5ccdbaa. Both implementation PRs are merged. Independent final Spec and Standards reviews approved.

Current notes

• PB-001 found no existing line-of-sight Gate in targeted cast admission. Add the appropriate shared Gate with supported exemptions and client/bot parity coverage so the movement prerequisite responds to a real typed Refusal. Channel actions remain unsupported until their running lifecycle is correlated.
• The historical Actor cast bypasses the player spellbook check. PB-003 owns the appropriate shared Gate for the new runner and must coordinate with provisioning in PB-005. An unlearned heal must refuse without cast, power, or cooldown changes.
• Started in isolated core and Package worktrees from merged PB-002. Shared runner composition belongs to PB-003; builds use one coordinated slot.
• Shared design keeps one durable objective and one foreground action. A moving leader refreshes the companion destination without replacing the retained heal identity. Each parallel branch must prove behavior against merged prerequisites.
• The companion recovery fixture must cover critical health at a reached leader destination. A permanent survival Hold cannot block both useful self-recovery and necessary bounded provisioning.
• The documented synchronous debug cast bypass is restored without weakening normal cast Gates. Migration now checks exact clean source trees, tools, Package content and each build artifact digest. Grouped non-healers retain self-recovery; mixed heal fixtures use the same class and role. Paired PR integration remains in progress.
• Final paired runs exposed two fixture assumptions. The retained-ally case must count observed completed movement legs, including the dispatch gap before a spline exists. Recovery revalidation now learns its spell explicitly after removing it. Failed runs remain attributed to their exact source; a corrected complete run is required before merge.

Add Warrior tank and Mage damage behavior beside the Priest healer on the shared runner. Make explicit durable roles, vanilla class capabilities, chosen-target assistance, controlled-enemy avoidance, death, rest, and regrouping work through repeated party pulls.

Acceptance criteria

• Warrior, Priest, and Mage have validated ability sets at every level supported by the companion fixture.
• Each companion retains an explicit tank, healer, or damage role, with a deterministic vanilla-derived default when no role was selected.
• The Warrior tank chooses and holds the party's selected fight according to survival and threat facts rather than inheriting damage-only retreat behavior.
• The Mage acts at class-appropriate range and role does not force a ranged class into melee.
• All companions honor the party's chosen fight and designated target and do not initiate an unrequested pull.
• Controlled enemies remain ineligible unless the control is intentionally broken by a higher-priority survival rule with a recorded reason.
• Repeated-pull fixtures cover wounded allies between fights, target death or disappearance, loss of line of sight, companion death, rest, and regrouping without role abandonment.
• Between fights, companions select useful missing class buffs, repair casting position when needed, apply them through real core casts, observe the resulting auras, and avoid redundant recasting while the buff remains valid.

Verified criteria: 8/8.

Warrior, Priest, and Mage have validated ability sets at every level supported by the companion fixture.

Reviewed by Orchestrator Spec review and independent Standards review.

• playerbots_level_five_roles_obey_classic_level_gates_and_never_pull_from_selection proves the level-5 Warrior spellbook and ordinary level Gates, then level-10 capability outcomes.
• The complete 87-case push run34317255372 passed against actual Core6f65ec9c and reviewed Packagefc72b93a, now merged as4c883874. The orchestrator verified the clean source, Wasm and Package identities for all68 case inputs and188 input/outcome records.

Each companion retains an explicit tank, healer, or damage role, with a deterministic vanilla-derived default when no role was selected.

Reviewed by Orchestrator Spec review and independent Standards review.

• The capability case and playerbots_populated_pb006_state_upgrades_roles_without_replacing_operator_catalogue prove explicit defaults and retained custom roles through a populated upgrade.
• The complete 87-case push run34317255372 passed against actual Core6f65ec9c and reviewed Packagefc72b93a, now merged as4c883874. The orchestrator verified the clean source, Wasm and Package identities for all68 case inputs and188 input/outcome records.

The Warrior tank chooses and holds the party's selected fight according to survival and threat facts rather than inheriting damage-only retreat behavior.

Reviewed by Orchestrator Spec review and independent Standards review.

• playerbots_tank_repairs_range_and_completes_a_real_taunt proves approach, timed Taunt completion and the Warrior becoming the creature threat target.
• The complete 87-case push run34317255372 passed against actual Core6f65ec9c and reviewed Packagefc72b93a, now merged as4c883874. The orchestrator verified the clean source, Wasm and Package identities for all68 case inputs and188 input/outcome records.

The Mage acts at class-appropriate range and role does not force a ranged class into melee.

Reviewed by Orchestrator Spec review and independent Standards review.

• The Mage engagement and repeated-pull casting-position records prove spell use at class range and repair after losing line of sight.
• The complete 87-case push run34317255372 passed against actual Core6f65ec9c and reviewed Packagefc72b93a, now merged as4c883874. The orchestrator verified the clean source, Wasm and Package identities for all68 case inputs and188 input/outcome records.

All companions honor the party's chosen fight and designated target and do not initiate an unrequested pull.

Reviewed by Orchestrator Spec review and independent Standards review.

• Selection-only and designated-engaged records prove that role selection does not start a pull and that the current eligible chosen fight wins.
• The complete 87-case push run34317255372 passed against actual Core6f65ec9c and reviewed Packagefc72b93a, now merged as4c883874. The orchestrator verified the clean source, Wasm and Package identities for all68 case inputs and188 input/outcome records.

Controlled enemies remain ineligible unless the control is intentionally broken by a higher-priority survival rule with a recorded reason.

Reviewed by Orchestrator Spec review and independent Standards review.

• playerbots_selected_engaged_target_wins_and_control_holds_every_role proves stun, polymorph, fear, root and pending control exclude damage until control clears.
• The complete 87-case push run34317255372 passed against actual Core6f65ec9c and reviewed Packagefc72b93a, now merged as4c883874. The orchestrator verified the clean source, Wasm and Package identities for all68 case inputs and188 input/outcome records.

Repeated-pull fixtures cover wounded allies between fights, target death or disappearance, loss of line of sight, companion death, rest, and regrouping without role abandonment.

Reviewed by Orchestrator Spec review and independent Standards review.

• playerbots_buffs_and_repeated_pulls_retain_roles_through_rest_los_and_death covers between-fight healing, target death and disappearance, line of sight, Mage death, resurrection and regroup. The two corrected provisioning-resume cases also pass against the merged Package.
• The complete 87-case push run34317255372 passed against actual Core6f65ec9c and reviewed Packagefc72b93a, now merged as4c883874. The orchestrator verified the clean source, Wasm and Package identities for all68 case inputs and188 input/outcome records.

Between fights, companions select useful missing class buffs, repair casting position when needed, apply them through real core casts, observe the resulting auras, and avoid redundant recasting while the buff remains valid.

Reviewed by Orchestrator Spec review and independent Standards review.

• The retained-ally buff, stronger-family and repeated-pull cases prove range and line-of-sight repair, real Fortitude/Frost Armor/Battle Shout auras, retained targets and no redundant recast.
• The complete 87-case push run34317255372 passed against actual Core6f65ec9c and reviewed Packagefc72b93a, now merged as4c883874. The orchestrator verified the clean source, Wasm and Package identities for all68 case inputs and188 input/outcome records.

Execution model: gpt-5.6-sol, high effort. Repositories: Official Package Collection starter class and role behavior, LyraCore Module combat and capability facts, Durable repeated-pull party scenarios.

Recorded evidence

• Started from merged companion Core 25c4bd9 and Package 284c0ab in isolated worktrees. The owner is implementing Warrior tank and Mage damage behavior, party target choice, controlled-enemy avoidance, repeated pulls, and real buff casts.
• Latest source review: Core f78f40d1a43e8d75089aef4615e6820f5af4ae8a and Package 3c6deab0ecad71303da0e1cd753f89c5440e5509. Prior runtime at Core 2aae1a1 passed 4 of 8 focused cases. Pre-upgrade evidence now names the actual merged PB-006 sources and Wasm.
• Focused Core f78f40d1a43e8d75089aef4615e6820f5af4ae8a with Package 3c6deab0ecad71303da0e1cd753f89c5440e5509: 8 passed, 2 failed in 138.55 seconds. Corrected test source b20b33b7f772d6c0887e3563b300ecc4d87b80dd is under verification. No merge or completed-ticket claim yet.
• Combined run at Core b20b33b7f772d6c0887e3563b300ecc4d87b80dd and Package 3c6deab0ecad71303da0e1cd753f89c5440e5509: roles 10/10, quest catalog 11/11, companion 18/20, provisioning 7/8, runner rewrite 36/37. The runner failure was a missing predecessor artifact setting. The other failures exposed a retained-heal-target regression, a seeded training gap, and an old refusal-name assertion. Full logs and failed states are preserved.
• Final focused run at Core 0e35ef28c5c9dd448068ab43e1d181b20ed7d819 and Package 5bd494a2d46a7b60f956c50d1d6050e1053f9ca0 passed all ten role cases in 124.46 seconds. The orchestrator verified 82 clean current-source records and two separately attributed PB-006 pre-upgrade records. The full-union release Wasm has SHA-256 84c29866c6b7cf8c24c557c563e8467d4cfa5a6c90331d70a034325b673750d4.
• Core durable CI run 34307984533 passed at the final reviewed head. The orchestrator verified every required check, the resolved review thread and unchanged base before the normal squash merge at 2026-09-09T03:59:46Z.
• Package PR CI run 34309281082 passed all 86 durable cases. Push run 34309277962 passed 85 and failed one companion assertion that read a cast event after its one-second retention period. The companion fixture is being corrected to use the captured cast identity, durable outcome and actual healing.
• At Core 2a72f91b576ea0912183edcebc15a100200a1678 and Package d0d3b491b0bd7b1c35976715190650e919613136, both focused interruption and retained-heal cases passed. The following local suite accidentally used parallel test threads: roles passed 11/11 and companions passed 16/20. That run does not meet the required serialized fixture conditions. The unchanged sources are being tested again with one test thread, and the failed run is preserved.
• Final serialized local run at Core 2a72f91b576ea0912183edcebc15a100200a1678 and Package d0d3b491b0bd7b1c35976715190650e919613136 passed all eleven role cases in 139.50 seconds and all twenty companion cases in 140.51 seconds. The orchestrator verified 109 clean current-source records and two PB-006 predecessor records. Current Wasm SHA-256 is 9d90933ac9724df89deca62425670a281f87ccf8a1a0dcbe4cac07222b8f0e50. Both final source reviews are clear.
• Core correction PR #493 passed every CI check and merged normally at 2026-09-09T05:05:33Z as 2b78c172e39fd479aeb477c1857768e9bf30120c. Its tree exactly matches reviewed Core 2a72f91. Package correction d0d3b491 is pushed to PR #30; final push run 34313526235 and PR run 34313530662 now exercise all 87 regression cases against the merged Core.
• At Package d0d3b491 and merged Core 2b78c172, final PR run 34313530662 passed all 87 durable cases and 1,637 library tests. Push run 34313526235 passed 86 of 87. Its retained-heal artifact proves cast 1 stayed pending across leader movement, resolved once, and raised health from 30/120 to 92/120. A later decision correctly started another heal because health remained below the 80% threshold. The fixture incorrectly expected the latest action-kind record to remain cast 1. Core PR #494 controls the observed decision phases and satisfies healing before checking resumed Follow. No production change is needed for this finding.
• Core #494 passed the focused retained-heal case in 11.84 seconds, both independent reviews and every CI check. The orchestrator verified all five saved behavior phases and the clean source/Wasm identities, then merged normally at 2026-09-09T06:00:31Z as 6f65ec9c1998170d36c2c4d76867e85718548619. The merge tree exactly matches reviewed head 417bdc9c. Final Package CI must now pass against this merged Core.
• Package #30 final head fc72b93a has the same reviewed source tree as d0d3b491. Push run 34317255372 and PR run 34317258868 are exercising the complete 87-case suite against merged Core 6f65ec9c. Final results and artifact verification remain pending.
• Final push run 34317255372 passed all 87 cases at Package fc72b93a and actual Core 6f65ec9c. The orchestrator verified 68 clean case inputs, 188 input/outcome records, the reviewed source tree, predecessor attribution and all required module checks. Package #30 merged normally at 2026-09-09T06:29:45Z as 4c883874854e64d942d85938533ed45b9bd725b2. No approval override was used.
• The parallel PR run passed 86/87 and missed a provisioning pass before Follow resumed. Its durable history proves Skill54 Applied1, then an expected spell139 Level Refusal and Follow. Core #495 now requires a new Applied outcome in the history, treating an absent initial row as zero. A Refusal alone cannot satisfy the test. Both focused resume cases passed at final Core 6e969ba3 and merged Package 4c883874. The correction has clear Spec and Standards reviews; Core CI is pending.
• Core #495 passed every check, including durable run34319791075, and merged normally at2026-09-09T06:56:15Z asfd6d230df3a75f642de36ceb4d56fb594aa3215d. The orchestrator verified the unchanged source/base, no open review threads, focused effect records, and a merge tree exactly equal to reviewed6e969ba3. All eight role criteria are now linked to reviewed evidence.

Current notes

• PB-004 is complete. Both repositories contain the role implementation, and the independent Core fixture corrections are merged. Earlier failed runs remain in the execution record.

Reconcile the supported bot's skills, spells, vanilla talents, equipment, bags, food, potions, ammo, reagents, and bounded supplies on spawn and level-up through real core operations. Make each action idempotent and explain it by provisioning profile and revision.

Acceptance criteria

• Provisioning runs on spawn and level-up and periodically repairs only resources owned by the selected profile.
• Spell learning, equipment, inventory, uniqueness, proficiency, class, and level rules pass through core operations. Free provisioning is an explicit profile choice; its grants remain bounded and attributable.
• Skill training and vanilla talent choices follow the selected class profile through the owning core Gates. Prerequisites, earned talent points, rank limits, and level restrictions hold, and reconciliation cannot spend a point or train a rank twice.
• Repeated reconciliation does not create duplicate equipment, unlimited supplies, repeated spell grants, or repeated costs.
• The policy respects an equipped upgrade, does not delete quest items, and never raises a bot's level to match a leader.
• Quest credit, earned XP, loot entitlement, and quest completion remain consequences of normal gameplay operations.
• Full bags, missing consumables, resource recovery, rest, and death produce bounded next actions or an explained stop.
• Decision history records the provisioning profile, revision, chosen action, and typed outcome.

Verified criteria: 8/8.

Provisioning runs on spawn and level-up and periodically repairs only resources owned by the selected profile.

Reviewed by Independent Spec and Standards source reviewers; orchestrator verified clean runtime artifacts.

• playerbots_provisioning_arms_levelup_but_control_modes_block_gameplay and playerbots_provisioning_arms_then_reconciles_and_repairs_without_cost prove arming, eligible reconciliation and periodic repair.
• Final Core-main CI 34292649436 passed 20 companion, eight provisioning and 37 shared durable cases, plus 1,634 Module library tests. Core 4353fb7786e7bb5e8afa6c600804cae69d316d8f and Package 1a482be932fb63c24a572aeef2c7c21c45741168 are merged with their reviewed source trees. Both independent source reviews approved; the orchestrator verified 57 clean case inputs and 68 attribution records.
• https://github.com/LyraCoreProject/packages/actions/runs/34292649436

Spell learning, equipment, inventory, uniqueness, proficiency, class, and level rules pass through core operations. Free provisioning is an explicit profile choice; its grants remain bounded and attributable.

Reviewed by Independent Spec and Standards source reviewers; orchestrator verified clean runtime artifacts.

• playerbots_provisioning_bounds_imported_style_trainer_facts and the inventory-preservation cases exercise owning class, level, spell, item and bounded grant Gates.
• Final Core-main CI 34292649436 passed 20 companion, eight provisioning and 37 shared durable cases, plus 1,634 Module library tests. Core 4353fb7786e7bb5e8afa6c600804cae69d316d8f and Package 1a482be932fb63c24a572aeef2c7c21c45741168 are merged with their reviewed source trees. Both independent source reviews approved; the orchestrator verified 57 clean case inputs and 68 attribution records.
• https://github.com/LyraCoreProject/packages/actions/runs/34292649436

Skill training and vanilla talent choices follow the selected class profile through the owning core Gates. Prerequisites, earned talent points, rank limits, and level restrictions hold, and reconciliation cannot spend a point or train a rank twice.

Reviewed by Independent Spec and Standards source reviewers; orchestrator verified clean runtime artifacts.

• playerbots_talent_gates_are_atomic_and_use_global_tab_identity and playerbots_provisioning_stops_on_partial_talent_catalog_without_character_changes prove shared eligibility, point accounting and bounded partial-catalogue refusal.
• Final Core-main CI 34292649436 passed 20 companion, eight provisioning and 37 shared durable cases, plus 1,634 Module library tests. Core 4353fb7786e7bb5e8afa6c600804cae69d316d8f and Package 1a482be932fb63c24a572aeef2c7c21c45741168 are merged with their reviewed source trees. Both independent source reviews approved; the orchestrator verified 57 clean case inputs and 68 attribution records.
• https://github.com/LyraCoreProject/packages/actions/runs/34292649436

Repeated reconciliation does not create duplicate equipment, unlimited supplies, repeated spell grants, or repeated costs.

Reviewed by Independent Spec and Standards source reviewers; orchestrator verified clean runtime artifacts.

• playerbots_provisioning_arms_then_reconciles_and_repairs_without_cost proves repeated reconciliation and bounded repair without duplicate grants or costs.
• Final Core-main CI 34292649436 passed 20 companion, eight provisioning and 37 shared durable cases, plus 1,634 Module library tests. Core 4353fb7786e7bb5e8afa6c600804cae69d316d8f and Package 1a482be932fb63c24a572aeef2c7c21c45741168 are merged with their reviewed source trees. Both independent source reviews approved; the orchestrator verified 57 clean case inputs and 68 attribution records.
• https://github.com/LyraCoreProject/packages/actions/runs/34292649436

The policy respects an equipped upgrade, does not delete quest items, and never raises a bot's level to match a leader.

Reviewed by Independent Spec and Standards source reviewers; orchestrator verified clean runtime artifacts.

• playerbots_provisioning_stops_cleanly_and_preserves_owned_items_and_gear proves preservation of upgrades and quest items.
• Final Core-main CI 34292649436 passed 20 companion, eight provisioning and 37 shared durable cases, plus 1,634 Module library tests. Core 4353fb7786e7bb5e8afa6c600804cae69d316d8f and Package 1a482be932fb63c24a572aeef2c7c21c45741168 are merged with their reviewed source trees. Both independent source reviews approved; the orchestrator verified 57 clean case inputs and 68 attribution records.
• https://github.com/LyraCoreProject/packages/actions/runs/34292649436

Quest credit, earned XP, loot entitlement, and quest completion remain consequences of normal gameplay operations.

Reviewed by Independent Spec and Standards source reviewers; orchestrator verified clean runtime artifacts.

• The reviewed profile uses core training and item operations without awarding XP, changing level, quest credit or quest completion. Durable snapshots preserve those gameplay facts.
• Final Core-main CI 34292649436 passed 20 companion, eight provisioning and 37 shared durable cases, plus 1,634 Module library tests. Core 4353fb7786e7bb5e8afa6c600804cae69d316d8f and Package 1a482be932fb63c24a572aeef2c7c21c45741168 are merged with their reviewed source trees. Both independent source reviews approved; the orchestrator verified 57 clean case inputs and 68 attribution records.
• https://github.com/LyraCoreProject/packages/actions/runs/34292649436

Full bags, missing consumables, resource recovery, rest, and death produce bounded next actions or an explained stop.

Reviewed by Independent Spec and Standards source reviewers; orchestrator verified clean runtime artifacts.

• playerbots_provisioning_yields_to_recovery_then_supplies_missing_recovery and the explicit-stop case prove bounded supplies, real item consumption, aura-driven recovery and explained stops.
• Final Core-main CI 34292649436 passed 20 companion, eight provisioning and 37 shared durable cases, plus 1,634 Module library tests. Core 4353fb7786e7bb5e8afa6c600804cae69d316d8f and Package 1a482be932fb63c24a572aeef2c7c21c45741168 are merged with their reviewed source trees. Both independent source reviews approved; the orchestrator verified 57 clean case inputs and 68 attribution records.
• https://github.com/LyraCoreProject/packages/actions/runs/34292649436

Decision history records the provisioning profile, revision, chosen action, and typed outcome.

Reviewed by Independent Spec and Standards source reviewers; orchestrator verified clean runtime artifacts.

• Captured provisioning history records the selected profile and revision with the action and typed result. The partial-catalogue and trainer cases verify explicit ProfileLimit and other Refusal causes.
• Final Core-main CI 34292649436 passed 20 companion, eight provisioning and 37 shared durable cases, plus 1,634 Module library tests. Core 4353fb7786e7bb5e8afa6c600804cae69d316d8f and Package 1a482be932fb63c24a572aeef2c7c21c45741168 are merged with their reviewed source trees. Both independent source reviews approved; the orchestrator verified 57 clean case inputs and 68 attribution records.
• https://github.com/LyraCoreProject/packages/actions/runs/34292649436

Execution model: gpt-5.6-sol, high effort. Repositories: Official Package Collection provisioning policy, LyraCore Module skill, spell, talent, item, inventory, and equipment operations, Durable idempotency scenarios.

Recorded evidence

• Final Core-main CI 34292649436 passed 20 companion, eight provisioning and 37 shared durable cases, plus 1,634 Module library tests. Core 4353fb7786e7bb5e8afa6c600804cae69d316d8f and Package 1a482be932fb63c24a572aeef2c7c21c45741168 are merged with their reviewed source trees. Both independent source reviews approved; the orchestrator verified 57 clean case inputs and 68 attribution records.
• The tested PR merge 130596d93eff39464e0e5d0484059bb37e0edf0f and final Package merge share tree 08b805920d8153532d8316c322e2f3f95d285c00. Core merge tree a446563e0c1db045b9f789bb02fc0fa715b4c112 equals the reviewed Core tree.
• Module BLAKE3 is 1bc2193ef2d92dfd634a109f9ea48ebf9d4380c070242180e79a3703f606e548; Package Content Identity is 0e4c8a1cd9015052d768abff8dd1183ffbd737494d969ccc44cf714762aa9614. The final main run preserves both identities from the passing paired run.
• The separate combined quest-admission CI 34291439781 passed all 76 durable cases and 1,637 library tests. The orchestrator verified 99 clean records and 68 case inputs. It includes the final fixture parking corrections and bounded post-cast provisioning before actual Follow movement.
• Every acceptance criterion has a named durable case in the final handoff. Tests cover raw trainer and talent limits, atomic Gate refusal, full bags, bank-slot equipment refusal, preservation of stronger equipment and quest items, actual recovery consumption, repeatable repair, control restrictions and populated upgrade.
• All actual review threads were resolved. The cancelled duplicate push module check was rerun successfully before merge. Branch protection was preserved.

Current notes

• An eligible Cohort runner applies at most one provisioning operation per pass, with current Account admission and no Foreground Action. Spawn and level-up arm reconciliation.
• Ready and Recorded permit the chosen gameplay action to continue. Worked consumes the pass and preserves the objective. Resurrection, useful survival, healing, recovery and defense outrank routine upkeep.
• Profiles use core spell, skill, talent, item and equipment Gates. Grants are explicitly free. Reconciliation never changes level, grants XP or quest credit, or replaces equal or stronger equipment.
• The seeded Warrior profile can explain MissingResource for Sunder Armor 7386 because the normal seed lacks its spell header. The complete-profile fixture names only valid seeded content. This does not establish imported trainer or world readiness.
• All runtime evidence uses owned loopback fixtures with named seeded content and geometry. Imported-world, sustained load, Transfer and attended-client acceptance remain assigned to later tickets.

Build a versioned quest catalog for the initial Northshire and Elwynn proof from imported relations, objectives, spawns, loot sources, and actual enders. Admit a quest only when every required objective has a supported executor, and retain the chosen objective and destination.

Acceptance criteria

• The catalog records quest eligibility, each objective kind, target or drop source, actual ender, work area, destination evidence, and content revision.
• The initial fixture contains a named supported Northshire and Elwynn catalog with deterministic seeds for each starter class.
• Admission checks ordinary core eligibility and confirms an executor for every required objective before acceptance.
• Unsupported exploration, escort, scripted event, transport, and complex GameObject requirements remain explicit and cannot enter a futile loop. Mixed quests are admitted only when every required kind has an executor, including supported talk-only and simple GameObject steps.
• A live creature search chooses what can be acted on now but does not decide whether a quest destination exists in imported content.
• The retained objective stores the actual ender and catalog revision and survives a tactical combat interruption.
• If a held quest becomes unsupported, the bot records the missing capability and proceeds to another supported objective or a specific wait without repeated abandonment and reacceptance.

Verified criteria: 7/7.

The catalog records quest eligibility, each objective kind, target or drop source, actual ender, work area, destination evidence, and content revision.

Reviewed by Independent Spec and Standards source reviewers; orchestrator verified clean runtime artifacts.

• playerbots_catalog_is_named_versioned_and_shared_by_starter_classes records the named catalogue, content identity, actual enders, destination evidence and revisions.
• Final Core-main CI 34294328577 passed 20 companion, eight provisioning, 11 catalogue and 37 shared durable cases, plus 1,637 Module library tests. Core dae05c1b78cfd4b59f24968ec8be41f75ce896d1 and Package 2028771d15c8efe87b50d2b6795315221953055d are merged with their reviewed trees. Both independent source reviews approved; the orchestrator verified 68 clean case inputs and 99 attribution records.
• https://github.com/LyraCoreProject/packages/actions/runs/34294328577

The initial fixture contains a named supported Northshire and Elwynn catalog with deterministic seeds for each starter class.

Reviewed by Independent Spec and Standards source reviewers; orchestrator verified clean runtime artifacts.

• playerbots_catalog_is_named_versioned_and_shared_by_starter_classes proves the Warrior, Priest and Mage fixture catalogue.
• Final Core-main CI 34294328577 passed 20 companion, eight provisioning, 11 catalogue and 37 shared durable cases, plus 1,637 Module library tests. Core dae05c1b78cfd4b59f24968ec8be41f75ce896d1 and Package 2028771d15c8efe87b50d2b6795315221953055d are merged with their reviewed trees. Both independent source reviews approved; the orchestrator verified 68 clean case inputs and 99 attribution records.
• https://github.com/LyraCoreProject/packages/actions/runs/34294328577

Admission checks ordinary core eligibility and confirms an executor for every required objective before acceptance.

Reviewed by Independent Spec and Standards source reviewers; orchestrator verified clean runtime artifacts.

• playerbots_talk_kill_creature_drop_and_provided_item_paths_change_core_state and playerbots_gameobject_loot_and_simple_use_paths_change_core_state prove admission followed by callable core executors.
• Final Core-main CI 34294328577 passed 20 companion, eight provisioning, 11 catalogue and 37 shared durable cases, plus 1,637 Module library tests. Core dae05c1b78cfd4b59f24968ec8be41f75ce896d1 and Package 2028771d15c8efe87b50d2b6795315221953055d are merged with their reviewed trees. Both independent source reviews approved; the orchestrator verified 68 clean case inputs and 99 attribution records.
• https://github.com/LyraCoreProject/packages/actions/runs/34294328577

Unsupported exploration, escort, scripted event, transport, and complex GameObject requirements remain explicit and cannot enter a futile loop. Mixed quests are admitted only when every required kind has an executor, including supported talk-only and simple GameObject steps.

Reviewed by Independent Spec and Standards source reviewers; orchestrator verified clean runtime artifacts.

• playerbots_admission_refuses_each_unsupported_requirement_and_mixed_work proves explicit unsupported and mixed-objective refusal.
• Final Core-main CI 34294328577 passed 20 companion, eight provisioning, 11 catalogue and 37 shared durable cases, plus 1,637 Module library tests. Core dae05c1b78cfd4b59f24968ec8be41f75ce896d1 and Package 2028771d15c8efe87b50d2b6795315221953055d are merged with their reviewed trees. Both independent source reviews approved; the orchestrator verified 68 clean case inputs and 99 attribution records.
• https://github.com/LyraCoreProject/packages/actions/runs/34294328577

A live creature search chooses what can be acted on now but does not decide whether a quest destination exists in imported content.

Reviewed by Independent Spec and Standards source reviewers; orchestrator verified clean runtime artifacts.

• playerbots_imported_destination_and_live_target_are_separate_facts proves a missing live target does not erase a static destination.
• Final Core-main CI 34294328577 passed 20 companion, eight provisioning, 11 catalogue and 37 shared durable cases, plus 1,637 Module library tests. Core dae05c1b78cfd4b59f24968ec8be41f75ce896d1 and Package 2028771d15c8efe87b50d2b6795315221953055d are merged with their reviewed trees. Both independent source reviews approved; the orchestrator verified 68 clean case inputs and 99 attribution records.
• https://github.com/LyraCoreProject/packages/actions/runs/34294328577

The retained objective stores the actual ender and catalog revision and survives a tactical combat interruption.

Reviewed by Independent Spec and Standards source reviewers; orchestrator verified clean runtime artifacts.

• playerbots_quest_objective_survives_combat_and_refreshes_changed_evidence proves retained purpose and refreshed content evidence.
• Final Core-main CI 34294328577 passed 20 companion, eight provisioning, 11 catalogue and 37 shared durable cases, plus 1,637 Module library tests. Core dae05c1b78cfd4b59f24968ec8be41f75ce896d1 and Package 2028771d15c8efe87b50d2b6795315221953055d are merged with their reviewed trees. Both independent source reviews approved; the orchestrator verified 68 clean case inputs and 99 attribution records.
• https://github.com/LyraCoreProject/packages/actions/runs/34294328577

If a held quest becomes unsupported, the bot records the missing capability and proceeds to another supported objective or a specific wait without repeated abandonment and reacceptance.

Reviewed by Independent Spec and Standards source reviewers; orchestrator verified clean runtime artifacts.

• playerbots_held_unsupported_quest_selects_supported_work_without_reaccepting proves alternative work or a specific wait, including loss or banking of a provided item.
• Final Core-main CI 34294328577 passed 20 companion, eight provisioning, 11 catalogue and 37 shared durable cases, plus 1,637 Module library tests. Core dae05c1b78cfd4b59f24968ec8be41f75ce896d1 and Package 2028771d15c8efe87b50d2b6795315221953055d are merged with their reviewed trees. Both independent source reviews approved; the orchestrator verified 68 clean case inputs and 99 attribution records.
• https://github.com/LyraCoreProject/packages/actions/runs/34294328577

Execution model: gpt-5.6-sol, high effort. Repositories: Official Package Collection quest catalog and admission, LyraCore imported quest and world relations, Seeded durable catalog scenarios.

Recorded evidence

• Final Core-main CI 34294328577 passed 20 companion, eight provisioning, 11 catalogue and 37 shared durable cases, plus 1,637 Module library tests. Core dae05c1b78cfd4b59f24968ec8be41f75ce896d1 and Package 2028771d15c8efe87b50d2b6795315221953055d are merged with their reviewed trees. Both independent source reviews approved; the orchestrator verified 68 clean case inputs and 99 attribution records.
• Tested PR merge a5de9bb7dac00bea2f0ac926903d5902878285d7 and final Package merge share tree c27a460ab7cdcdd2072a90aa3275bc68e53b15c0. Core merge tree 432809ddefbc666d8dfe9ccde288eca0488bd8c7 equals the reviewed and tested source.
• Module BLAKE3 is a7af4aeb2bda1ea386841e3312701d66ccd06b11bbf04092d3a2b9631f57c77e; Package Content Identity is 573c5856e3a261e073a7bd414e2183efa7f6dcec823a47d0c190d29b1aab3db4. Both identities agree with the preceding passing paired CI.
• The final source preserves all eight Package behavior patches through rebase. The only final workflow change restores current Core checkouts to main. The preceding full local composition also passed all 76 cases.
• The real deferral case waits through 30 seconds and retries without replacing the quest purpose. Human-led party control preserves the held quest and proves actual Follow movement after bounded upkeep. All three finite-objective expiry paths retain their intended behavior.
• Private staging refuses existing semantic rows or import records before writes. Identical owned staging is repeatable; changed revisions or later imports refuse. Typed loot storage plans preserve atomic inventory refusal and Gateway compatibility.
• All actual review threads were resolved before merge. Required duplicate push checks passed before optional duplicate work was cancelled. Final merge preserved branch protection.

Current notes

• The named twelve-quest Northshire/Elwynn catalogue records ordinary eligibility, all required objective executors, destinations, work areas, actual enders and separate content identity.
• Quest 3904 is chest-backed item collection. The private direct GameObject-use variant proves a separate executor and does not redefine the source quest.
• PB-007 owns ordinary available-quest acceptance, autonomous sequencing and earned rewards. The admission fixture can stage prerequisites and cannot establish an autonomous chain.
• The current fixture-only live creature finder must become a bounded current-partition read before production use. Recurring held-quest selection must account for rewarded history beyond the active quest limit.
• All runtime evidence uses owned loopback fixtures with named seeded content and geometry. Imported-world and attended-client gates remain open. The orchestrator is building the exact merged predecessor for later schema changes.

Complete supported kill, collect, talk-only, and simple GameObject quest loops. The bot accepts the quest, retains its objective, travels through real navigation, executes the required gameplay, earns ordinary credit and XP, turns in to the actual ender, and selects the next supported work.

Acceptance criteria

• A supported kill quest completes through accepted combat, actual damage and death, kill credit, travel, and turn-in rather than a changed internal label.
• A supported collect quest hunts an explicit eligible drop source, receives the item through real loot and inventory operations, and turns in to the catalog's actual ender.
• A supported talk-only chain advances through actual quest eligibility, giver and ender interactions, and ordinary rewards without manufacturing prerequisite completion.
• At least one cataloged simple GameObject objective completes through the owning core use or loot operation, respects range and object state, and earns credit once. Unsupported scripted object behavior is refused at admission.
• Quest objective progress changes only when the authoritative core gameplay outcome changes it.
• The retained quest objective survives combat interruption, death, rest, and provisioning, then resumes after rechecking eligibility.
• Target death, disappearance, or ineligibility cancels only the current action and selects a new eligible target or an explicit respawn wait.
• A completed turn-in grants ordinary rewards once and advances to another supported catalog entry without Operator help.
• Seeded scenarios cover Warrior, Priest, and Mage and produce reproducible decision and outcome histories.
• A supported quest destination outside the former home leash remains reachable under the current work area and retained destination. Tactical retreat returns to an appropriate safe position in the current partition and cannot reset travel to the spawn home.

Verified criteria: 10/10.

A supported kill quest completes through accepted combat, actual damage and death, kill credit, travel, and turn-in rather than a changed internal label.

Reviewed by Orchestrator Spec review and independent Standards review; final canonical evidence checked by orchestrator.

• `playerbots-quest-loop-{warrior,priest,mage}-*-{warrior,priest,mage}.json` begins with ordinary acceptance of Quest 7, travels more than 150 yards, records real admitted attacks or resolved casts, observes ten dead targets and ten Core credits, returns to the actual ender, and records one ordinary reward.
• Canonical 103-case run: https://github.com/LyraCoreProject/packages/actions/runs/34337451062 ; artifact10099537951. Merged Corec600a7cc and Packagea17e5ec7 match the reviewed source.

A supported collect quest hunts an explicit eligible drop source, receives the item through real loot and inventory operations, and turns in to the catalog's actual ender.

Reviewed by Orchestrator Spec review and independent Standards review; final canonical evidence checked by orchestrator.

• `playerbots-quest-loop-mage-*-mage-collect-reward.json` records owning requirement 8, eight real loot receipts, peak carried item count 8, ordinary turn-in once, rewarded Quest 33, and carried count 0 after reward consumption. `playerbots-quest-loop-gameobject-alternative-*` proves entitled GameObject loot skips a depleted nearer chest and uses the available chest more than nine yards away.
• Canonical 103-case run: https://github.com/LyraCoreProject/packages/actions/runs/34337451062 ; artifact10099537951. Merged Corec600a7cc and Packagea17e5ec7 match the reviewed source.

A supported talk-only chain advances through actual quest eligibility, giver and ender interactions, and ordinary rewards without manufacturing prerequisite completion.

Reviewed by Orchestrator Spec review and independent Standards review; final canonical evidence checked by orchestrator.

• The three class records begin unearned, accept and turn in talk Quest 783 through real giver and ender interactions, then proceed to Quest 7. Durable `on_quest_turnin` rows distinguish actual rewards from latest action observations.
• Canonical 103-case run: https://github.com/LyraCoreProject/packages/actions/runs/34337451062 ; artifact10099537951. Merged Corec600a7cc and Packagea17e5ec7 match the reviewed source.

At least one cataloged simple GameObject objective completes through the owning core use or loot operation, respects range and object state, and earns credit once. Unsupported scripted object behavior is refused at admission.

Reviewed by Orchestrator Spec review and independent Standards review; final canonical evidence checked by orchestrator.

• `playerbots-quest-loop-simple-gameobject-*-simple-gameobject.json` and `*-busy.json` prove range-aware Core use, one credit, a busy-object refusal, selection of an available alternative, and ordinary turn-in. The 3904 loot chest case remains separate.
• Canonical 103-case run: https://github.com/LyraCoreProject/packages/actions/runs/34337451062 ; artifact10099537951. Merged Corec600a7cc and Packagea17e5ec7 match the reviewed source.

Quest objective progress changes only when the authoritative core gameplay outcome changes it.

Reviewed by Orchestrator Spec review and independent Standards review; final canonical evidence checked by orchestrator.

• Kill counts, loot receipts, carried items, dead state, health, XP, money, and rewarded rows come from Core operations. `playerbots-quest-loop-record-only-*-record-only.json` records unchanged Character, quest, item, object, cast, spline, and action snapshots across RecordOnly passes.
• Canonical 103-case run: https://github.com/LyraCoreProject/packages/actions/runs/34337451062 ; artifact10099537951. Merged Corec600a7cc and Packagea17e5ec7 match the reviewed source.

The retained quest objective survives combat interruption, death, rest, and provisioning, then resumes after rechecking eligibility.

Reviewed by Orchestrator Spec review and independent Standards review; final canonical evidence checked by orchestrator.

• `playerbots-quest-loop-survival-*` records actual incoming damage, Survival precedence, retreat to a visited safe position, lethal damage and death, and unchanged Quest purpose. `playerbots-quest-loop-upkeep-*` records a real retained heal that raises authoritative health, then Provisioning while the quest identity, sequence, and zero credit remain unchanged.
• Canonical 103-case run: https://github.com/LyraCoreProject/packages/actions/runs/34337451062 ; artifact10099537951. Merged Corec600a7cc and Packagea17e5ec7 match the reviewed source.

Target death, disappearance, or ineligibility cancels only the current action and selects a new eligible target or an explicit respawn wait.

Reviewed by Orchestrator Spec review and independent Standards review; final canonical evidence checked by orchestrator.

• `playerbots-quest-loop-disappearance-*` and `*-ineligible-target-*` retain Quest purpose while the bounded selector replaces only the missing or refused target through the owning attack Gate. The raw corpse and corpse-prefix cases distinguish Missing from ReadLimit. GameObject cases replace depleted or busy sources.
• Canonical 103-case run: https://github.com/LyraCoreProject/packages/actions/runs/34337451062 ; artifact10099537951. Merged Corec600a7cc and Packagea17e5ec7 match the reviewed source.

A completed turn-in grants ordinary rewards once and advances to another supported catalog entry without Operator help.

Reviewed by Orchestrator Spec review and independent Standards review; final canonical evidence checked by orchestrator.

• Durable turn-in receipts remain one for Quests 783, 7, 5261, 33, 3904, and the simple-use quest after later work. Rewarded rows, XP and money changes, and later retained catalogue entries prove forward progress without duplicate rewards.
• Canonical 103-case run: https://github.com/LyraCoreProject/packages/actions/runs/34337451062 ; artifact10099537951. Merged Corec600a7cc and Packagea17e5ec7 match the reviewed source.

Seeded scenarios cover Warrior, Priest, and Mage and produce reproducible decision and outcome histories.

Reviewed by Orchestrator Spec review and independent Standards review; final canonical evidence checked by orchestrator.

• The three class records contain separate navigation, runner, action, Core outcome, quest, reward, and content identities. Priest starts with declared 400 mana and resolves real Smite 585 using source-derived cost, level, school, family, and effect fields plus declared fixture choices for its missing auxiliary DBC values.
• Canonical 103-case run: https://github.com/LyraCoreProject/packages/actions/runs/34337451062 ; artifact10099537951. Merged Corec600a7cc and Packagea17e5ec7 match the reviewed source.

A supported quest destination outside the former home leash remains reachable under the current work area and retained destination. Tactical retreat returns to an appropriate safe position in the current partition and cannot reset travel to the spawn home.

Reviewed by Orchestrator Spec review and independent Standards review; final canonical evidence checked by orchestrator.

• The class records reach the retained Quest work area more than 150 yards from spawn. Survival uses an out-of-combat position that the Character actually visited in the current partition. `playerbots-quest-loop-safe-partition-*-safe-partition-change.json` proves that retained point is invalidated on partition change.
• Canonical 103-case run: https://github.com/LyraCoreProject/packages/actions/runs/34337451062 ; artifact10099537951. Merged Corec600a7cc and Packagea17e5ec7 match the reviewed source.

Execution model: gpt-5.6-sol, high effort. Repositories: Official Package Collection autonomous quest behavior, LyraCore Module quest, combat, loot, inventory, and navigation operations, Seeded durable quest scenarios.

Recorded evidence

• Clean release build passed in 1 minute 32 seconds at Core 8bb5637190ea7e2dd9375e28136c8a4e6502d0b2 and Package 0e41d1e605fb60f32ab6cfcf784f43f9643f2ad4. This proves compilation only. No quest-loop runtime acceptance is claimed.
• Initial focused diagnostic at Core 7330536424d4e450ae8b97bcc4211bc95631ac4f and Package 0e41d1e605fb60f32ab6cfcf784f43f9643f2ad4 stopped after failures. Zero of twelve cases are accepted. The retained timeout records Quest 7 at zero credit, no movement progress, and repeated Hold decisions. No quest-loop acceptance is claimed.
• Second provisional release build passed in 1 minute 35 seconds at Core 0ea7818ab2ab897e4f3c5397159060089c685028 and Package d6287eea6899c9e728c6d779af897a363f5e1d46. Ordinary native compilation then failed on five generated references to debug-only fixture registrations. None of the fourteen current quest cases ran in this attempt.
• Corrected provisional run at Core d46397b80644a832394da4b58773860b67e152e3 and Package 67af06ce9ab79e5cff673f960750c05d7208f16e completed fourteen cases: eight passed and six failed in 654.71 seconds. The release Wasm build, ordinary native build and three registry-boundary tests passed. Twelve saved current-source records match the clean source pair, Wasm and Package Content Identity. The complete failed run is preserved.
• Focused corrections passed the chest collection, RecordOnly, simple GameObject and survival cases. The populated PB-004 upgrade then reached a real retained cast and exposed a schema layout incompatibility: a new quest-action payload increased the persisted Action size from 24 to 32 bytes. Reordering the never-merged payload fields preserves the original size and typed target identity. The corrected populated upgrade is running again.
• Corrected provisional populated upgrade passed in 9.15 seconds and recovery/provisioning passed in 50.97 seconds at Core bb8e7e11c41425fcbd4b423bc2bb3ceeb8a1fb8f and Package 95169639bd7252724272d5c0efeebf2de09dd7b8. The orchestrator verified saved Wasm hashes, both clean current-source records, the exact retained cast and runner, and every existing quest-objective field. New fields defaulted to null. Later test source now also records the actual health gained during recovery; that addition awaits the final run.
• The clean provisional composition at Core 173609f70b597630fa25744532c10b4235ebb9be and Package f2c6104ca9702cbf5091597e86f62842eb4e5dfa passed 13 of 14 cases in 538.43 seconds. The retreat test staged damage while the Character still occupied its observed safe position. Corrected Core 0c612e90 and Package b4610e35 wait for actual movement away from the current safe position, then compose real incoming damage and one normal runner decision atomically. Focused and complete reruns are pending. The populated upgrade and health-backed recovery/provisioning cases passed in the 13-case result.
• The corrected interruption/death case passed in 41.96 seconds and the corrected survival/read-limit case passed in 38.19 seconds at Core 0c612e90 and Package b4610e35. The following full rerun exposed a separate Mage observation timeout. Its saved state shows all twelve supported quests already rewarded, including the required collect quest, and a completed return home with no failures. The test waited for a temporary inventory count after the bot had already turned those items in. The owner is correcting the evidence capture; the full run remains unaccepted.
• Final receipt correction is reviewed at Core 6f68b15d6e7bf9bfbf37c2cf75f66da03e93b92e and Package 37b4a6f373bc4ff3c3b83ebdbf768aead211176b. The fixture records actual loot notifications, received quantity, peak carried quantity and source. The Mage case requires eight received item750, a peak of eight, the real collect requirement, consumed inventory and exactly one ordinary turn-in. It permits the bot to advance to later valid work between reads. The corrected runtime run remains pending.
• At Core f2da7c2fdbcfaf7a131af502f69b722e7a167618 and Package 37b4a6f373bc4ff3c3b83ebdbf768aead211176b, the release and native builds passed, all 31 Package API checks passed, the corrected all-class case passed in 299.07 seconds, and survival/read-limit passed in 37.98 seconds. The complete fourteen-case run is still active. Current Wasm SHA-256 is 8996b32713c3b3ca56860262ee0c249d17d422911bb4d5ef7c30d4ecbbf3d4f1.
• The final provisional quest suite passed all fourteen cases in 544.73 seconds at Core f2da7c2f and Package 37b4a6f. The orchestrator verified the saved Wasm, 22 clean current-source runtime records, real collect requirement and eight-item receipt, consumed inventory and once-only turn-in. Source work is now rebased onto actual merged Package 4c883874. The actual role predecessor and final 101-case integrated run remain required.
• Final replay is clear on both review axes at Core 2e082c64 and Package 952f1c49. Every changed quest source blob matches the fourteen-case passing composition. The workflow now builds the actual immediate role predecessor from merged Core fd6d230d and Package 4c883874. The populated upgrade is in progress; final combined 101-case CI remains required.
• The actual populated predecessor upgrade passed one of one cases in 9.00 seconds. Root verified exact retained cast, runner and existing quest fields; both new fields defaulted to none. Core PR496 is open. CI caught two unsupported generated index-column declarations; the owner restored the generated binding and the standard Rust checks now pass. Durable CI remains active.
• Core PR496 merged normally at 07:42 UTC on September 9 after all checks passed and both review axes cleared. Actual merge 6ab7a2c1 has the exact reviewed source tree. Package PR31 is open at 168ffe3e and pins that merged Core for the final 101-case suite. PB-007 completion remains pending that integrated result and Package merge.
• Package PR31 canonical run34325525771 at 168ffe3e completed with 97 of 101 cases passing. All fourteen quest-loop cases passed. Four older cases failed on companion selection, unsupported quest completion, retained quest safe-position changes and defense setup. Their saved states and full run are preserved. The owner is reviewing minimal fixture corrections; the final integrated run and Package merge remain required.
• Final review found two additional production issues before Package merge: active held quests could fall outside the first sixteen catalog entries, and refused quest interactions lacked the existing candidate retry delay. The owner is adding a bounded twenty-slot active quest read with explicit overflow handling and reusing the existing scaled retry delay. A separate GameObject map/instance correction affects only debug fixture staging. The final combined total is now 102 cases; none of these corrections is accepted until its runtime evidence passes.
• The final correction sources passed thirteen focused behavior cases: all twelve catalog cases and the Defense regression. The final held-quest case used clean Core aee33c35 and Package c59513ce; the unaffected cases used the earlier test-only predecessor with identical Package source. Release compilation, native compilation and all 31 Package API checks passed. Both review axes and default-Core binding generation precede the follow-up Core PR; the actual merged Core and complete 102-case canonical Package CI remain required.
• The last review correction passed at clean Core df0dc006 and Package 0fb17393. A changed quest interaction now starts at retry count one with a one-second delay. Active quest-log overflow no longer extends an existing Defense retry deadline; its real-damage boundary case passed with normal one-second eligibility. Release and native builds, all 31 Package API checks and both focused cases passed. Both review axes are clear. The follow-up Core PR and final 103-case canonical Package CI remain required. Saved Wasm SHA-256 4baf7975e7e70ca12c39d8ef52e0e89ad6629bca75a789d334256f6a0e2b2361.
• Follow-up Core PR497 is open at c1d0c2f4. Rust formatting, native build, Wasm build, Clippy and tests passed. Durable binding CI and refreshed review remain pending. The only source differences from the final focused passing pair are three formatting changes, which the orchestrator verified. Final 103-case Package CI follows the actual Core merge.
• Core PR497 merged normally as c600a7cc3887508012a512c77d7fefc128cf791d. Its tree exactly matches the reviewed head. Package PR31 now pins that merge at head8bd60167. The required push module check passed, all three inline review threads have replies and are resolved, and canonical 103-case run34337451062 is active. Package completion remains pending that result and refreshed review.
• Package PR31 merged normally at 10:32 UTC on September 9 as a17e5ec7120de21fe9ccc75c3b8fcc998be35df8. Its tree exactly matches reviewed head8bd60167. Canonical run34337451062 passed all103 durable cases and all four predecessor builds. Root independently verified69 clean case-input records and220 clean input/outcome records at exact Corec600a7cc and Package8bd60167, the recorded source tree, one current Wasm identity, the full CI log hash, and all three resolved review threads. No configured realm or client acceptance is claimed.

Current notes

• Complete with all103 canonical durable cases, all four predecessor builds, the actual populated upgrade and both review axes clear. Earlier failed attempts remain preserved. These seeded fixtures do not replace the imported-world and attended-client release observations.
• The implementation now shares the role ticket's bounded combat-spell selector. PB-004 is therefore an explicit integration dependency. Source work can proceed in parallel through the agreed provisional composition; PB-007 cannot complete before the corrected role code merges.

Turn typed failure and progress evidence into finite recovery for both autonomous bots and companions. Change approach, defer destinations, invalidate stale targets, and explain waits without erasing the retained objective.

Acceptance criteria

• The blocked-target fixture cannot clear its stall clock through attack acceptance when route, damage, and quest progress remain unchanged.
• After 10 seconds without movement progress the bot changes approach; within 30 seconds it advances, defers the destination, or records a specific wait.
• A safe partial route remains eligible while it advances a route or waypoint within a bounded route budget.
• Repeated no-progress defers a destination by map, instance, and geometry revision for a bounded period and allows another useful action.
• Unknown coverage stays distinct from a verified blocked route and reports the missing imported coverage evidence.
• Death, full bags, absent targets, an unreachable giver, a controlled enemy, and a broken prerequisite each end in recovery or an explained finite stop.
• Recovery remains active for companions and uses follow distance, leader movement, and casting position as its progress evidence.
• Restarting the fixture preserves the bounded failure memory that prevents immediate reselection of the same failed target.

Verified criteria: 8/8.

The blocked-target fixture cannot clear its stall clock through attack acceptance when route, damage, and quest progress remain unchanged.

Reviewed by Independent Standards review and orchestrator Spec, source-attribution and merge review.

• The ordinary-scheduler blocked-target case preserves unchanged target health, position and quest credit through repeated attack admission, then changes approach and selects another eligible fight.
• Canonical run https://github.com/LyraCoreProject/packages/actions/runs/34353096953 passed all 122 exact expected cases. Artifact 10106472761 contains 88 clean input records and 240 current source records. Both PRs merged normally after passing checks and resolved review threads.

After 10 seconds without movement progress the bot changes approach; within 30 seconds it advances, defers the destination, or records a specific wait.

Reviewed by Independent Standards review and orchestrator Spec, source-attribution and merge review.

• The stalled-attack case observes approach change after ten seconds and deferral at the thirty-second threshold with scheduler tolerance. The original repeated-acceptance failure and all correction attempts remain saved.
• Canonical run https://github.com/LyraCoreProject/packages/actions/runs/34353096953 passed all 122 exact expected cases. Artifact 10106472761 contains 88 clean input records and 240 current source records. Both PRs merged normally after passing checks and resolved review threads.

A safe partial route remains eligible while it advances a route or waypoint within a bounded route budget.

Reviewed by Independent Standards review and orchestrator Spec, source-attribution and merge review.

• The partial-route case observes actual westward movement on a bounded 4096-expansion Partial route toward an eastern destination. Recovery retains the route while its owned legs advance.
• Canonical run https://github.com/LyraCoreProject/packages/actions/runs/34353096953 passed all 122 exact expected cases. Artifact 10106472761 contains 88 clean input records and 240 current source records. Both PRs merged normally after passing checks and resolved review threads.

Repeated no-progress defers a destination by map, instance, and geometry revision for a bounded period and allows another useful action.

Reviewed by Independent Standards review and orchestrator Spec, source-attribution and merge review.

• The failed-work cases prove partition and navigation identity, finite deferral, useful healing and exact-target retry after actual navigation import. The active-coverage case proves same-generation effective coverage growth advances the revision and permits retry.
• Canonical run https://github.com/LyraCoreProject/packages/actions/runs/34353096953 passed all 122 exact expected cases. Artifact 10106472761 contains 88 clean input records and 240 current source records. Both PRs merged normally after passing checks and resolved review threads.

Unknown coverage stays distinct from a verified blocked route and reports the missing imported coverage evidence.

Reviewed by Independent Standards review and orchestrator Spec, source-attribution and merge review.

• The verified-blocked geometry case records Blocked without missing coverage. The Unknown case records MissingImportedCoverage and NoMovement. Both pass in the canonical suite.
• Canonical run https://github.com/LyraCoreProject/packages/actions/runs/34353096953 passed all 122 exact expected cases. Artifact 10106472761 contains 88 clean input records and 240 current source records. Both PRs merged normally after passing checks and resolved review threads.

Death, full bags, absent targets, an unreachable giver, a controlled enemy, and a broken prerequisite each end in recovery or an explained finite stop.

Reviewed by Independent Standards review and orchestrator Spec, source-attribution and merge review.

• Canonical recovery, failure, companion and catalog cases cover actual respawn, full inventory with real loot and once-only reward, absent GameObject, unreachable actual ender, controlled quest during Survival, bounded unknown reads and finite prerequisite-cycle admission.
• Canonical run https://github.com/LyraCoreProject/packages/actions/runs/34353096953 passed all 122 exact expected cases. Artifact 10106472761 contains 88 clean input records and 240 current source records. Both PRs merged normally after passing checks and resolved review threads.

Recovery remains active for companions and uses follow distance, leader movement, and casting position as its progress evidence.

Reviewed by Independent Standards review and orchestrator Spec, source-attribution and merge review.

• Moving-leader and real self-heal cases pass. The casting-position case proves eight owned movement legs while the same ally remains at 30 health, so movement clears the matching Heal attempt without unrelated regeneration.
• Canonical run https://github.com/LyraCoreProject/packages/actions/runs/34353096953 passed all 122 exact expected cases. Artifact 10106472761 contains 88 clean input records and 240 current source records. Both PRs merged normally after passing checks and resolved review threads.

Restarting the fixture preserves the bounded failure memory that prevents immediate reselection of the same failed target.

Reviewed by Independent Standards review and orchestrator Spec, source-attribution and merge review.

• The persistent-process case stops and restarts its owned Standalone, retains exact Recovery, deferrals, objective and quest, then prevents immediate reselection of the same failed target.
• Canonical run https://github.com/LyraCoreProject/packages/actions/runs/34353096953 passed all 122 exact expected cases. Artifact 10106472761 contains 88 clean input records and 240 current source records. Both PRs merged normally after passing checks and resolved review threads.

Execution model: gpt-6-astra, high effort. Repositories: Official Package Collection travel and recovery behavior, LyraCore Module navigation evidence, Durable blocked-route and stalled-work scenarios.

Recorded evidence

• Source review found that the existing no-movement branch records failure and repeats the same movement request. Attacks at an already reached quest destination can also repeat without entering that movement branch. The first recovery fixture will require a changed approach or finite deferral when accepted attacks produce no actual movement, damage or quest progress. Runtime proof is pending.
• The first live diagnostic reproduced the defect at clean Core 6578b7cdab76a6533ab08efbd282e0ab2f540953 and Package 338578f98d354e34f1d9e2a7d829d8a33bc0e650. Across 36 samples and 34.243 seconds, the ordinary scheduler repeatedly accepted attacks on one blocked Quest 7 target. Position, target health and quest credit stayed unchanged. No approach change, failure record or deferral occurred. The required behavior test failed as expected. The saved Wasm SHA-256 is 356b6ef1c2758a57c8770883bfaa14fe250624ae54df984a2ec71f9fbcb4c56c.
• The first recovery slice passed its ordinary-scheduler case in 41.18 seconds at Core 24986740fb04ea546de457e70630811bef2b2f60 and Package 50ddce397e22a86680a440c3f5d512c22d0a5445. Across 36 saved samples, the bot selected a different approach at 10.646 seconds and deferred the blocked target at 31.156 seconds, within the declared two-second observation tolerance. Actual position, target health and quest credit remained unchanged. Current Wasm SHA-256 is 93b45a2fe39ef0fd5f85eb4cd6aec9c2808103380f2e07dadffb47e3f3c1f02f.
• The second slice passed two of three durable cases in 91.49 seconds at clean Core b7da940aaac1cacf863ff621078a27f016c17b1c and Package cff2d24f441a11c0bd639e6765c6c8b4a4cef90d. Blocked quest recovery and moving-leader recovery with a real self-heal passed. The persistent fixture failed on a missing setup row before restart; restart acceptance is unproved. Wasm SHA-256 f31bc070cbf52c1215ba719610b754920eeb18de8f5454f7225330fd26df338e. A preceding build stopped on a disposable-cache disk quota and is preserved separately.
• The third slice passed 117 pure Package checks and two of six durable cases in 141.18 seconds at clean Core 26b23706526988c7bf53699bb32ae3dcbb63e760 and Package 52c8c5d69ffce32337a94076d02b5b3a003df31a. Blocked quest recovery selected another eligible fight, and moving-leader recovery allowed a real self-heal. Restart and navigation-import setup failed because the text reader discarded an empty single-column row. The controlled-enemy case exposed a missing failure record. The partial-route fixture had gaps in its wall and moved toward the leader. These failures remain preserved; the fourth run is in progress after corrections. Wasm SHA-256 3daedb0d8287b5ae5399f2ef6bbb6e94e1bf328d2f0a3083635e2fabacbd6bd4.
• The fourth slice passed all seven recovery cases in 181.14 seconds and 117 pure Package checks at Core 0aa0ca6e9c63e1e70d6814fadc265a41f69fc869 and Package bdf892ee2a2b082cf5dd99e12417f8fa1e7ecf72. Root independently verified ten clean input records, saved Wasm, exact failure-memory preservation across a real process restart, movement away from the leader on a Partial route, and navigation-import rollback. The unchanged tactical-progress regression passed. The old blocked-home test still expected whole-objective deferral; the quest retry setup tried to exhaust an attempt before provisioning had allowed one to start. Both failures are preserved and corrected source awaits verification. Wasm SHA-256 23b81cdfd3e7ce94f986c00d3b44d70adb2d69cde59fd96f2c4378e5be25afa5.
• Default-Core Gateway generation passed at Core 9a446ae6. Exactly the navigation-revision table, type and registration changed; all 1002 Gateway binding files match the generator. Core c529dd53 and Package 8ec7668f are frozen for ten recovery scenarios and two corrected existing regressions. The populated migration was independently reviewed but awaits the actual merged quest Package predecessor.
• At frozen Core c529dd539efa14f3f21821f62bf058cf6aa97b52 and Package 8ec7668f35fc91662b0493e01894f726e65e73ca, 117 pure checks and nine of ten durable recovery scenarios passed in 212.73 seconds. The controlled-survival test lacked an observed safe position, so the runner correctly selected ControlledHold. The corrected setup now observes safe ground and actual movement before a real incoming hit. It remains unrun. The corrected blocked-home regression passed in 138.09 seconds including its nested release build. Root verified eleven clean input records and the saved Wasm, SHA-256 57cd21371c5b49074428c0d68a9019fc1404dc1dd58459c0e104ad6f5913b71b.
• At clean Core c3379ef0 and Package ace66264, release compilation, 117 pure checks and 31 API checks passed. Five of seven focused durable cases passed: full inventory recovery, unreachable actual quest ender, controlled quest during Survival, active quest overflow with Defense, and changed interaction retry. Respawn checked before real arrival; retained-quest deferral allowed real movement after staged exhaustion. Both setups are corrected and await a rerun. Root verified eleven clean source records, seven case identities and the saved Wasm, SHA-256 080e1c9f4b038e811c4e541fd81b6bf2ed698d7cd9859d1f103497144d5be3c5.
• Final independent review found that effective derived navigation coverage could grow without changing its generation id. Core c777c8ff now advances the existing navigation revision in the same transaction when new coverage becomes effective. Independent source review is clear. The current clean Core b91f50bc and Package 5b9ade15 include seven focused correction cases, including real Heal CastingPosition progress, verified blocked coverage and retry of the exact failed target after same-generation coverage growth. These cases are prepared and have not run.
• Seventh run at clean Core b91f50bc and Package 5d8a998 passed release compilation,117 policy checks,31 Package API checks and four of eight focused cases. Actual respawn, exact-target navigation import, same-generation coverage retrofit and populated PB007 upgrade passed. Root verified eight clean case-input records, eleven current input/outcome records, the exact predecessor and40 saved runtime files. All30 existing Runner fields and the existing quest, cast, provisioning and spell rows survived upgrade unchanged; Recovery defaulted to none. Saved Wasm SHA-2566403ce4b3e1bef4fad749c22542e58020f061535aa333dd318e76592c4d5e52c.
• The seventh run found a real quest-retention defect: when the quest GameObject disappears, movement cancels correctly but live destination admission discards the still-active retained quest. Its correction must retain prior admitted placement evidence only after every current capability Gate passes. The other three failures concern an obsolete retry-count expectation, a serialized blocked-status spelling and natural ally regeneration masking the casting-movement proof. Source corrections are in progress. Default-Core generation passed for all1002 binding files, and all1454 Gateway tests passed at Core b91f50bc.
• The final quest-retention correction checks every current capability before reusing the exact retained destination identity. A new boundary case removes both the quest object and its Core objective and must refuse that fallback. The casting fixture now uses actual rooted melee engagement to hold ally health constant while proving movement clears the same Heal attempt. Both corrections are source-reviewed and await runtime evidence.
• At clean Core42ece0a2 and Package6a778480, all 12 catalog cases and verified blockage passed, with 117 policy checks and 31 API checks. Root verified 16 clean input records, 45 current input/outcome records and 78 saved runtime files. Wasm SHA-2563359132206dc55557cf2e459ec4653fc7b7e4ce91585803d2a3dd6e6c2bebd73. Three failures concerned an early forced pass, a missing observation field and an unnecessary continuous-melee assertion. The quest stayed retained; ally health stayed constant through eight completed movement legs. The final corrected replay remains pending.
• The final focused run passed all three cases at clean Core dae45d92480eaca3877aaf991c50db8b16ea0d3d and Package 6a778480bb284ed06c8d09f437231a68ea9618cd. Both missing-GameObject boundaries passed. The Priest completed eight owned movement legs across 12.160 seconds while ally health stayed at 30 in all nine samples; matching Heal recovery retained zero stall time. Root verified three clean input records, thirteen runtime files and the exact Wasm, SHA-256 3359132206dc55557cf2e459ec4653fc7b7e4ce91585803d2a3dd6e6c2bebd73. The prior casting failure remains saved: a distant Root caster added threat and allowed evade followed by regeneration. The corrected fixture uses the nearby ally as caster. Core PR #498 is open; canonical Package acceptance and both normal merges remain pending.
• Core #498 merged normally after all required checks passed at 7cb687e5 and both review threads were resolved. The merge is a880bd29b9a56f6fe685f858eb90bb748623dfa7. Final test corrections normalize empty lists and sort fixture GUIDs. Package #32 pins this merged Core and is running the full 122-case suite; recovery remains in review until that suite and the paired merge finish.
• Package #32 now includes the requested reproduction instructions at 4c5e30b1. CodeRabbit passed and its documentation thread is resolved. Canonical run 34353096953 has passed all five predecessor builds, Module compilation and Script Artifact checks. The 122-case durable suite is running. The README change advances Package content identity to eb76b3853a6b868a64a3d59fd05fe9017d3d52faeedc7ec999b6d105ed66fcdb.
• Recovery is complete. Core #498 merged as a880bd29b9a56f6fe685f858eb90bb748623dfa7 and Package #32 as a3d67002184d2ca427ffef8fd3e2504fe4d4ed1d. All 122 canonical cases passed. Root verified exact case names, 88 clean input records, 240 current source records and Package identity eb76b3853a6b868a64a3d59fd05fe9017d3d52faeedc7ec999b6d105ed66fcdb. The populated PB007 upgrade preserved all 30 existing Runner fields plus quest, cast, provisioning and spell state. Recovery defaulted to none.

Current notes

• Recovery source work starts from reviewed provisional Core 0c612e90 and Package b4610e35 in separate worktrees. This overlaps implementation with role and quest verification. PB-004 and PB-007 remain mandatory integration dependencies; final PRs must use their actual merged sources and prove populated upgrade plus the full regression suite. No acceptance criterion is waived.
• Seven durable recovery scenarios and the unchanged tactical-progress regression have passed. The current run adds unknown control reads, survival during a controlled quest and cancellation of an obsolete recovery leg. Final populated upgrade, all combined regressions and review remain required.
• Recovery now composes with the reviewed active-quest read limit and scaled interaction retry. A bounded read pause cannot start a recovery approach or accrue stalled-work time. Separate source work adds full-bag, all-targets-absent and unreachable-giver scenarios before recovery acceptance. The corrected controlled-survival setup and final composition still need runtime evidence.
• Independent review found that movement toward an unavailable quest GameObject could remain armed. The correction cancels that exact owned movement and retains the quest purpose. The new cancellation case and both corrected fixture setups are source-reviewed. Runtime, actual populated predecessor upgrade and the final combined suite remain required.
• Current recovery source is Core30a6ce50 and Package6a778480. All 12 catalog cases and verified blockage passed. The remaining test corrections preserve exact quest retention, real restored-object movement and constant-health casting progress. A disk quota and one test compile error stopped the next replay before runtime; both are corrected. The next run compiles its test callers first, then builds the Wasm and runs those three cases. Final canonical acceptance remains 122 cases.
• Source review is clear after the final corrections. The full 122-case Package suite must use the actual Core #498 merge. Passing focused seeded cases does not supply imported-world or attended-client acceptance.
• Core CI passed its unit test job but reported test-only lint findings. The sample assertions were extracted without changing predicates, and the saved loot-source identifier now compares by numeric value. One redundant borrow is corrected locally. A complete local lint pass is running before the next push; the final Module and Package behavior remains unchanged.
• Final canonical acceptance supersedes the earlier pending-run notes. The preceding failures remain part of the evidence history. Imported-world and attended-client acceptance belong to PB-011 and PB-012 and remain pending.

Complete the companion command contract through the existing command transport. Persist follow, stay, assist, and designated-target orders, apply them with party authority, and resume valid solo work when human-led control ends.

Acceptance criteria

• A human party leader can issue follow, stay, assist, and designated-target orders through the existing command transport.
• Each accepted order persists as typed control state and produces an explanation entry with Actor, target where relevant, and outcome.
• Follow tracks the leader, stay holds the commanded location within movement tolerances, and assist limits initiation to the chosen fight.
• A designated target participates in candidate identity and cannot be replaced by another proposal's target when scores change.
• Controlled enemies remain ineligible while assist or target orders are active unless an explicit higher-priority survival decision records why control was broken.
• Joining a human-led party suspends autonomous travel and opportunistic grouping without deleting the solo objective.
• Leaving the party invalidates companion actions, rechecks the retained solo objective, and resumes it only when it is still eligible.
• Role and command state survive companion death and remain consistent after regrouping.

Verified criteria: 8/8.

A human party leader can issue follow, stay, assist, and designated-target orders through the existing command transport.

Reviewed by Independent Standards review and orchestrator Spec, runtime attribution and merge review.

• The authenticated Follow, Stay, named-member Assist and exact Target cases pass through the command transport. Gateway cases cover durable command delivery, issuer fencing and restart. Forged ownership and reclaimed Account cases refuse the command.
• Canonical run https://github.com/LyraCoreProject/packages/actions/runs/34405533396 passed 139 Module cases and six Gateway cases. The orchestrator checked 325 current source records and three predecessor records in artifact 10126986818. Core e025ab7696f557ff10a88a1bb33710e3066aedc3 and Package 7858c7fdd6d0d4dd49b431419360e8ed22d69388 are merged. The reviewed head, tested merge ref and final Package merge have the same tree, 23218e2b0d570061a0a916e68829b4849a46b64e.

Each accepted order persists as typed control state and produces an explanation entry with Actor, target where relevant, and outcome.

Reviewed by Independent Standards review and orchestrator Spec, runtime attribution and merge review.

• Accepted orders retain typed state, issuer GUID, revision and receipt history. Follow repetition records Unchanged without restarting the retained cast. History rollover and capacity cases preserve durable receipts.
• Canonical run https://github.com/LyraCoreProject/packages/actions/runs/34405533396 passed 139 Module cases and six Gateway cases. The orchestrator checked 325 current source records and three predecessor records in artifact 10126986818. Core e025ab7696f557ff10a88a1bb33710e3066aedc3 and Package 7858c7fdd6d0d4dd49b431419360e8ed22d69388 are merged. The reviewed head, tested merge ref and final Package merge have the same tree, 23218e2b0d570061a0a916e68829b4849a46b64e.

Follow tracks the leader, stay holds the commanded location within movement tolerances, and assist limits initiation to the chosen fight.

Reviewed by Independent Standards review and orchestrator Spec, runtime attribution and merge review.

• The companion and order suites prove leader movement, Stay restrictions on ordinary and recovery positioning, in-range healing, and Assist against only the named member's actual fight.
• Canonical run https://github.com/LyraCoreProject/packages/actions/runs/34405533396 passed 139 Module cases and six Gateway cases. The orchestrator checked 325 current source records and three predecessor records in artifact 10126986818. Core e025ab7696f557ff10a88a1bb33710e3066aedc3 and Package 7858c7fdd6d0d4dd49b431419360e8ed22d69388 are merged. The reviewed head, tested merge ref and final Package merge have the same tree, 23218e2b0d570061a0a916e68829b4849a46b64e.

A designated target participates in candidate identity and cannot be replaced by another proposal's target when scores change.

Reviewed by Independent Standards review and orchestrator Spec, runtime attribution and merge review.

• The Target case checks the exact eligible creature and invalidates controlled work. Candidate identity and role regression cases retain the selected action, target, spell and objective when priorities change.
• Canonical run https://github.com/LyraCoreProject/packages/actions/runs/34405533396 passed 139 Module cases and six Gateway cases. The orchestrator checked 325 current source records and three predecessor records in artifact 10126986818. Core e025ab7696f557ff10a88a1bb33710e3066aedc3 and Package 7858c7fdd6d0d4dd49b431419360e8ed22d69388 are merged. The reviewed head, tested merge ref and final Package merge have the same tree, 23218e2b0d570061a0a916e68829b4849a46b64e.

Controlled enemies remain ineligible while assist or target orders are active unless an explicit higher-priority survival decision records why control was broken.

Reviewed by Independent Standards review and orchestrator Spec, runtime attribution and merge review.

• The named-member Assist and exact Target cases retain current control Gates. Companion and role regressions cover controlled enemies, safe alternatives and recorded Survival behavior.
• Canonical run https://github.com/LyraCoreProject/packages/actions/runs/34405533396 passed 139 Module cases and six Gateway cases. The orchestrator checked 325 current source records and three predecessor records in artifact 10126986818. Core e025ab7696f557ff10a88a1bb33710e3066aedc3 and Package 7858c7fdd6d0d4dd49b431419360e8ed22d69388 are merged. The reviewed head, tested merge ref and final Package merge have the same tree, 23218e2b0d570061a0a916e68829b4849a46b64e.

Joining a human-led party suspends autonomous travel and opportunistic grouping without deleting the solo objective.

Reviewed by Independent Standards review and orchestrator Spec, runtime attribution and merge review.

• The retained solo Quest case observes a real party join, a companion objective and an unchanged retained Quest. Sessionless consent and companion regressions cover suspension of opportunistic grouping.
• Canonical run https://github.com/LyraCoreProject/packages/actions/runs/34405533396 passed 139 Module cases and six Gateway cases. The orchestrator checked 325 current source records and three predecessor records in artifact 10126986818. Core e025ab7696f557ff10a88a1bb33710e3066aedc3 and Package 7858c7fdd6d0d4dd49b431419360e8ed22d69388 are merged. The reviewed head, tested merge ref and final Package merge have the same tree, 23218e2b0d570061a0a916e68829b4849a46b64e.

Leaving the party invalidates companion actions, rechecks the retained solo objective, and resumes it only when it is still eligible.

Reviewed by Independent Standards review and orchestrator Spec, runtime attribution and merge review.

• The retained solo Quest case proves actual departure, inactive companion order, cleared Follow work and a rechecked Quest under a new runner identity. Catalog and recovery regressions refuse stale destinations and unsupported work.
• Canonical run https://github.com/LyraCoreProject/packages/actions/runs/34405533396 passed 139 Module cases and six Gateway cases. The orchestrator checked 325 current source records and three predecessor records in artifact 10126986818. Core e025ab7696f557ff10a88a1bb33710e3066aedc3 and Package 7858c7fdd6d0d4dd49b431419360e8ed22d69388 are merged. The reviewed head, tested merge ref and final Package merge have the same tree, 23218e2b0d570061a0a916e68829b4849a46b64e.

Role and command state survive companion death and remain consistent after regrouping.

Reviewed by Independent Standards review and orchestrator Spec, runtime attribution and merge review.

• The death case retains the Follow order and Tank role while choosing Resurrection. Companion resurrection, leadership-change and retained-solo-Quest cases verify recovery and consistent roles after membership changes.
• Canonical run https://github.com/LyraCoreProject/packages/actions/runs/34405533396 passed 139 Module cases and six Gateway cases. The orchestrator checked 325 current source records and three predecessor records in artifact 10126986818. Core e025ab7696f557ff10a88a1bb33710e3066aedc3 and Package 7858c7fdd6d0d4dd49b431419360e8ed22d69388 are merged. The reviewed head, tested merge ref and final Package merge have the same tree, 23218e2b0d570061a0a916e68829b4849a46b64e.

Execution model: gpt-5.6-sol, high effort. Repositories: Official Package Collection commands and control state, LyraCore Gateway command transport and Module party reads, Durable join and leave scenarios.

Recorded evidence

• The command owner has separate Core and Package worktrees on feat/playerbots-companion-orders. The reviewed design covers authenticated command bytes, Realm-core party authority, durable delivery receipts, bounded retries and follow/stay/assist/target behavior.
• The first two full-union compile attempts exposed new serialized-enum and generated-API declaration errors. The owner corrected them; no command runtime acceptance is claimed. Current source is Core 98ad4cce and Package 69fa51b2, queued for the next serialized verification slot.
• Release Wasm compilation and default-Core Gateway binding generation passed. Subsequent Gateway compilation found generated lookup and test-fixture type errors. The fifth attempt at Core 070f229c and Package 9ffeddb0 stopped on fixture parse types and an unavailable artifact-hash dependency. Each failed build is preserved. No companion-command runtime acceptance is claimed.
• At clean Core a6f63ab3 and provisional Package 87eec9d0, the full-union release Wasm compiled in 1 minute 36 seconds and Gateway tests compiled in 2 minutes 26 seconds. Focused checks are now running. The branch includes retained solo-quest join/order/leave and populated quest-predecessor upgrade scenarios; their runtime acceptance is pending.
• Gateway authority and encoded command transport checks passed at Core a6f63ab3 and Package 87eec9d0. The partial durable run passed receipt-capacity and forged-issuer cases, then recorded five failures before it was stopped. Independent review found five production defects: cache-only receipt decisions could duplicate commands across Transfer, Assist omitted target-loss outcomes, relay workers and scanning lacked a fair bound, final application lacked the full body and consent Gate, and Realm admission did not certify the exact party roster. The owner is correcting these source paths. No command runtime acceptance is claimed.
• Further command review found a Coordinator lock held across network work, an expired queue head that could block later commands, and delayed command delivery after Transfer finish that could create conflicting receipts. The owner has source corrections and real Transfer boundary scenarios in progress. Source receipt removal now occurs explicitly before Transfer cleanup. An indeterminate outcome after receipt retention expires remains visible as OutcomeUnknown. These changes are not yet accepted by runtime evidence.
• The corrected command source entered build validation at Core 0e12485d and Package 864090d1. Review found two remaining behavioral defects: Gateway ignored StalePartyMirror from exact roster certification, and switching command issuers could erase the fence against an older retry. Both corrections and their boundary cases remain required before acceptance.
• Default-Core command binding generation passed. The current corrected source entered the full-union release build at Core22af6253 and Package539cf2f0; a displaced inline API exemption then stopped the Package firewall before Rust compilation. Restoring its exact merged form produced Packagef4a3c7b6 and the owner resumed validation. Source review clears stale authority handling and the per-issuer command fence. A whole-cache membership scan still needs an indexed read before final acceptance.
• At clean Core c26c803f and Package f4a3c7b6, default-Core binding generation, Gateway compilation and the keyed membership, stale mirror and roster-overflow checks passed. The full-union release build passed in 1 minute 42 seconds. The issuer-Transfer fixture then published a default Core artifact after its Package wrapper had exited and stopped because playerbots_spawn was absent. That setup failure is preserved. The owner is rerunning with the Package wrapper retained through the complete fixture process; command runtime acceptance remains pending.
• The original two-case replay reached actual Transfer import and failed on a Character uniqueness collision. Its leadership-cycle case failed during party setup with StaleWorldSession. Neither case proved command ordering. Source corrections now use disjoint topology-aligned identifiers, unique fixture Character names and authenticated party changes. They also preserve RecordOnly behavior when an Assist target is unavailable. Corrected runtime evidence remains pending.
• At clean Core 27d29797 and Package e92ff20f, the corrected intervening-leadership cycle passed in 6.96 seconds after a full-union release build. Issuer Transfer reached the destination but its next fixture step required a body that had not been created. Assist selected movement toward the exact target before the test prematurely expected melee. Both failed attempts remain saved; their setup corrections and final merged-predecessor composition are pending.
• The retry at clean Core 7194b23c and Package e92ff20f passed release and Gateway compilation. Issuer Transfer then failed at the Realm Account lookup during ordinary Session claim. Assist recorded movement toward the exact target, but the test searched for an uppercase movement tag. Core 7e3e04ca now adds exact Account and Session read checks and corrects the serialized tag. Source review is clear; the next replay remains pending. Both failures and identical prebuild/runtime Wasm hashes are preserved.
• At the retry-seven candidate, Assist passed in 16.55 seconds and Stay after Recovery passed in 18.82 seconds. The next issuer run passed the closed Claim, ordinary transferred Session, newer Applied command and delayed Superseded command. Its final reply assertion exposed the test SQL parser splitting a quoted payload at an embedded vertical bar. The owner is fixing that input parser; the issuer case is not yet counted as passed.
• The issuer Transfer and causal-fence case passed in 13.01 seconds at clean Core 8b7b04a05dc70c0f6a1588582efff6730422098d and Package 5b1849ca8882bd4503197dda311c34189a07bc5c. Both structured SQL parser cases and native Gateway compilation passed. The retained pre/post Wasm hashes match, SHA-256 19efd9e6a4369b2e81dd16ac5f502ba6850bd854a0219073523cd34ff67490fe. Assist and Stay passed at the preceding fixture revision. Final default-Core checks then found two architecture failures after 1523 Module tests passed. Four Character lookups and command-intent lifecycle classification are corrected at 6db23c86; a clean rerun is required before the PR opens.
• Final native checks at Core 85bebcd and Package e482898 passed both corrected Gateway cases, all 1352 Gateway tests and all 172 shared tests. The prior Module run passed all 1525 tests. Core 4145e3e5 adds only two function-local test length lint exceptions, cleared by root review. The exact Clippy rerun and normal PR checks remain pending.
• Core PR500 is open at b5a4fdb8. Build, format, Wasm, Clippy and general tests pass in CI; durable checks remain pending. CI exposed missing command-table parity coverage, then nested generated types that needed parity support. The corrected parity harness passed all 118 cases and target Clippy locally. Root reviewed the exact test-only changes. Earlier failed CI runs remain part of the record.
• PR500 review identified missing Gateway partition checks and missing ownership validation for transferred command-sequence rows. The owner is correcting those boundaries, serializing test environment changes, limiting repeated lane-overflow logs and returning the precise stale-party outcome. Two reports were checked against the code and runtime evidence: pending intents already use an unreapable deadline, and the questioned SQL projection already passed on SpacetimeDB 2.7.1. Merge waits for the corrected revision, checks and resolved review threads.
• All seven PR500 review threads are resolved at Core 014da9d8. The corrected source passed 71 party tests, all 118 schema parity checks and full workspace Clippy. The two issuer import tests and canonical duplicate-table test also passed. Two test commands that selected zero cases remain recorded and were replaced by the full named party suite. The corrected revision is now in CI; merge and the paired Package release remain pending.
• Core PR500 merged normally at 50018fcda9f22acdc8f66d1d14d8b91a8513e27a. The reviewed 014da9d8 revision passed all seven checks, including the 21-minute durable suite, and all seven review threads were resolved. The paired Package is being pinned to this actual merge for canonical CI; PB-009 remains incomplete until that release passes and merges.
• Package PR33 pins merged Core 50018fcd. Its first Module checks found four fixture reads that bypassed the Transfer fence. The reviewed correction uses fenced reads and a bounded private-party Account check. A further small correction preserves existing fight facts when a designated target is already present. Current Package 1629f783 is in canonical CI; the single review thread is resolved. Earlier failures remain recorded.
• Both canonical runs of Package1629f783 failed in Gateway fixtures. The capacity test compared expiring total message counts, one same-Shard test read the Gateway destination before its subscription caught up, and the crash test expected a Transfer refusal before opening Escrow. The owner is preparing a narrow Core test follow-up, then will pin its actual merge and repeat Package CI. No command Package merge is claimed.
• At unchanged Core4a30d080 and Package1629f783, Gateway native compilation, focused Clippy and release Wasm passed. A local user quota then stopped native fixture compilation before gameplay. After unused cache cleanup, the same-Shard command Transfer passed in 10.79 seconds. Capacity passed its exact no-ack check but later received IntentNotDispatchHead. The crash case passed real Escrow, Transfer, old-holder refusal and exact source-finalization checks, then failed on a later history command outcome. Both failures remain under diagnosis; no follow-up Core or Package merge is claimed.
• Source review clears Core e7fb3a50 with Package 1629f783. The capacity fixture rotates sixteen pending commands through the real claim and defer operations before retrying the original head. The crash fixture rolls order history and applies the distinct-source command while the target body exists, then transfers all receipts and verifies old receipt replay without changing the later order. The three focused durable cases are queued. No runtime result is claimed for this correction.
• All three focused command cases passed at Core 6ec663d7 and Package 1629f783: capacity recovery in 154.81 seconds including its Module build, same-Shard receipt Transfer in 10.94 seconds, and crash boundaries in 26.98 seconds. The orchestrator verified all twenty runtime records against the exact clean source pair and Wasm, the restored queue head, unchanged order after late replay, and distinct source receipts for the same numeric id. Core PR #501 now has diagnostic-only head 9ab5c30b and is in normal CI and review.
• Core PR #501 merged normally at 2f2b1f35dde90e52dd3b0d3e16d44bb3ffe41a2d after all seven checks passed at reviewed 9ab5c30b. Its merge tree matches the reviewed tree, and review is resolved. Package PR #33 now pins that actual merge at 5b380265; canonical run 34393198302 is active. PB-009 remains incomplete until the Package checks pass and it merges.
• Canonical Package #33 run 34393198302 at Package 5b380265 and Core 2f2b1f35 failed four of seventeen Module order cases. All six Gateway cases, the populated Recovery upgrade and the other Module targets passed. The four failed assertions sampled a buff, an approach or an eligibility delay before the intended action. Source review clears bounded ordinary-action waits and explicit existing mana and survival fixture inputs at Core fbccb8ec. The owner will run the order suite, merge the Core follow-up through normal review, then repin and repeat Package CI. Both failed CI runs remain saved.
• The focused retry at unchanged Core 7d84b90f and Package 5b380265 passed native compilation and release Wasm after a stale predecessor build cache was removed. The solo Quest case then failed after 35.95 seconds. Source and saved rows show the cause: its helper removes the companion fixture group 5090300, while this test uses role group 5098000. The owner will use the existing authenticated membership operation and record actual membership removal. No production behavior change or Package merge is claimed; other focused cases still need execution.
• At clean Core fe53ebc3 and Package 5b380265, native compilation passed in 1 minute 9 seconds and release Wasm in 2 minutes 13 seconds. All four corrected cases passed, followed by all 17 order cases in 158.29 seconds, including the populated merged-PB-008 upgrade. The orchestrator independently verified 79 runtime records against the exact source pair and saved Wasm. Current Wasm SHA-256 is 1b99f0d69937ab88443b901d34d48810a42915399d74e87de97d07057501be47. The Core follow-up PR, actual Package repin and canonical Package CI remain required.
• Core PR #502 merged normally at e025ab7696f557ff10a88a1bb33710e3066aedc3 after all checks passed at reviewed fe53ebc3 and no review threads remained. Package #33 is being rebased and pinned to this actual merge for canonical CI. Its merge remains required before PB-009 is complete.
• Commands are complete. Canonical run https://github.com/LyraCoreProject/packages/actions/runs/34405533396 passed 139 Module cases and six Gateway cases. The orchestrator checked 325 current source records and three predecessor records in artifact 10126986818. Core e025ab7696f557ff10a88a1bb33710e3066aedc3 and Package 7858c7fdd6d0d4dd49b431419360e8ed22d69388 are merged. The reviewed head, tested merge ref and final Package merge have the same tree, 23218e2b0d570061a0a916e68829b4849a46b64e. Package content identity is 7b51ae0d6ab39ae35b7c4368e86baf02b9f62f2da67e42ad669caa55e2b78642. The current CI Wasm has BLAKE3 742e5205550f5d4b6e3898fb839a4eb6e1eee2a9344ac08d3b6187b84dc01731. Both repositories passed their required checks and review threads were resolved before normal merges.

Current notes

• After the final role correction passed all 31 sequential role and companion cases, command source work starts from the verified Core 2a72f91 and Package d0d3b491 pair. PB-004 remains a mandatory integration dependency. Before command PRs, rebase onto its actual merged sources and prove the populated predecessor and full regression suite. This overlap does not waive any role or command acceptance gate.
• The owner is replacing cache-only decisions with acknowledged receipt reads and transporting the bounded receipt window with the Character. Application is fenced during Transfer. The other corrections cover exact party roster certification, full body and consent admission, typed Assist target loss, and bounded fair command dispatch. Source review and runtime remain pending.
• The final merge order is quests, recovery, commands, then Transfer. Commands must prove their populated migration against the actual merged Recovery state and show that Stay gates the final recovery movement. Source work continues in parallel.
• Fair retry dispatch must also preserve the human's latest command. Review found that an older Stay retry could overwrite a newer Follow. The owner is adding a Core-minted monotonic command sequence that travels with the issuer Character. The target order retains its sequence when inactive and refuses a superseded command. Existing receipt lookup still returns the original result for an already applied command. Real issuer-Transfer ordering evidence remains pending.
• A separate provisional composition includes current Recovery before command control. Its Stay Gate applies after Recovery chooses a movement candidate. New source-only cases cover Stay over RecoveryPosition and populated migration with nonempty Recovery state. The original command pair remains isolated for its corrected issuer-Transfer and intervening-leadership checks. Neither composition is accepted yet.
• The corrected original command pair remains frozen for issuer Transfer and intervening-leadership verification. Shared build metadata from another checkout stopped a prebuild; cleaning only that crate restored the release build. A missing optional hash command then stopped the evidence script before runtime. Both attempts are preserved and the same two cases are being rerun.
• The command owner now has the serialized private verification slot for corrected issuer Transfer, intervening leadership and RecordOnly/Assist boundaries. Independent source review and separate populated Recovery predecessor CI wiring are in progress. Root has no local Cargo or realm process.
• Both source review axes clear the corrected original pair at Core 7194b23c and Package e92ff20f. The issuer now completes ordinary arrival into the companion target partition before issuing its new command. Assist waits for the selected movement to finish before requiring melee. Their private replay is queued after Transfer foundation checks.
• The combined command run exposed three test defects. Release retains a closed Account Claim, a short real movement leg completed during evidence collection, and injured fixture allies caused healing to preempt blocked Follow. Both reviewers cleared the corrections at Core 80d5dfead1c47658c3b929469f3d3982146a5f09 and Package 6329140236a6889404cc0d828ca3539ddd128431. The new focused run is pending. Earlier failed runs remain saved.
• Core PR #502 has passed the orchestrator review at fe53ebc3 and is running normal CI. The command owner will merge only after exact-head checks and review threads are clear, then pin that actual merge in Package #33 and repeat canonical Package CI.
• Final canonical acceptance and the actual Package merge supersede earlier pending-CI notes. Original failures remain in the evidence history. Transfer and final release acceptance remain separate requirements.

Define and prove the lifecycle of every new bot row across Character deletion, Module restart, Gateway restart, party changes, and Transfer. Suspend local actions while Transfer is pending, rebuild transient observations on arrival, and prevent stale work from acting.

Acceptance criteria

• Each durable Package row has an explicit keep, move, rebuild, or delete rule for Character deletion and Transfer.
• A pending Transfer suspends local actions and relies on authoritative Transfer state rather than timeout to create another executor.
• The retained quest or party objective, control mode, role, deadlines, deferred destinations, and bounded failure memory survive their defined restart and Transfer cases.
• Transient targets and route segments do not Transfer; arrival rebuilds them against the destination's content and geometry revisions.
• Pending work carries controller generation and location, and an old completion cannot advance a new action after restart or arrival.
• Restart fixtures interrupt active movement, cast, provisioning, quest reward, and Transfer stages without duplicating a Character, item, supply grant, reward, or active controller.
• A transferred companion retains the correct party and role on arrival, and a transferred autonomous bot resumes or invalidates its objective for a recorded reason.
• One instance entry and exit through the real routing path preserves companion orders and produces arrival evidence.

Verified criteria: 8/8.

Each durable Package row has an explicit keep, move, rebuild, or delete rule for Character deletion and Transfer.

Reviewed by Codex orchestrator and independent Standards and Spec reviewers.

• The reviewed Package README defines deletion and Transfer for every Character-owned and Shard-owned production table. The populated predecessor upgrade and ownership Architecture Tests pass.
• Final Package2453f012 has tree790da2ca094c9af6506bed444f808a099f1ba223, identical to the fully tested PR head. Core503, Core504 and Core506 are merged.

A pending Transfer suspends local actions and relies on authoritative Transfer state rather than timeout to create another executor.

Reviewed by Codex orchestrator and independent Standards and Spec reviewers.

• The twelve-boundary real Gateway restart matrix proves one pending source fence and authoritative arrival release. The final six Gateway scenarios pass in canonical CI.
• All eight Module Transfer cases pass, including pending-intent leases, retained source control and stale completion. No timeout creates a second executor.

The retained quest or party objective, control mode, role, deadlines, deferred destinations, and bounded failure memory survive their defined restart and Transfer cases.

Reviewed by Codex orchestrator and independent Standards and Spec reviewers.

• The final Transfer job passes eight Transfer, four lifecycle, three order, five arrival Recovery and one populated-upgrade case. They retain controller, purpose, role, order, deadlines and bounded failure memory under the defined transitions.
• Compatible and incompatible Quest destination cases and selected-member Assist pass. The broader job also passes retained healing movement and Stay after Recovery.

Transient targets and route segments do not Transfer; arrival rebuilds them against the destination's content and geometry revisions.

Reviewed by Codex orchestrator and independent Standards and Spec reviewers.

• The two Quest destination cases use separate source and destination catalogues and Navigation Inputs. Compatible Quest7 resumes; incompatible content records TransferPurposeChanged before selecting eligible destination work.
• Assist rebuilds destination movement toward the selected Priest. The broader suite passes safe-position partition invalidation and GameObject disappearance followed by a rebuilt approach.

Pending work carries controller generation and location, and an old completion cannot advance a new action after restart or arrival.

Reviewed by Codex orchestrator and independent Standards and Spec reviewers.

• The Module Transfer cases prove generation advance, refusal of stale source cast completion and exact AreaTrigger replay. The Gateway restart matrix binds each boundary to one Intent, controller generation and location revision.
• The real entry and exit scenario reaches Realm location revision3 and leaves no stale source action after release. All these cases pass in final canonical CI.

Restart fixtures interrupt active movement, cast, provisioning, quest reward, and Transfer stages without duplicating a Character, item, supply grant, reward, or active controller.

Reviewed by Codex orchestrator and independent Standards and Spec reviewers.

• Four lifecycle cases pass interruption of active movement, provisioning, a cast and completed Quest work through real process restart or Module replacement. They compare durable items, supplies, rewards and control ownership.
• The populated upgrade, twelve Gateway abort boundaries and arrival mirror repair pass. Final canonical CI reruns every corresponding Transfer case and all 146 broader regression cases.

A transferred companion retains the correct party and role on arrival, and a transferred autonomous bot resumes or invalidates its objective for a recorded reason.

Reviewed by Codex orchestrator and independent Standards and Spec reviewers.

• Assist and party mirror repair preserve the exact role and selected member, certify both World mirrors from Realm facts and resume destination movement.
• The Quest destination scenarios prove compatible resumption and a recorded TransferPurposeChanged replacement. The final instance route preserves the authenticated Follow order and party through both crossings.

One instance entry and exit through the real routing path preserves companion orders and produces arrival evidence.

Reviewed by Codex orchestrator and independent Standards and Spec reviewers.

• The real Gateway route enters map36 through trigger78 and exits through trigger119 to map0. It retains the authenticated Follow order, objective identity and deadline, advances the Realm location to revision3 and matches both World mirrors to Realm.
• The local route passes in36.81seconds at Core73617467 and Package2a8b3223. Final canonical CI repeats the route successfully at merged Corec147b0e4 and Packageffb14b42 before the normal Package merge.

Execution model: gpt-6-astra, high effort. Repositories: Official Package Collection bot row lifecycle, LyraCore Module, Gateway, Realm-core, Transfer, and Group Intent integration, Durable restart and Transfer scenarios.

Recorded evidence

• Foundation checks passed at clean Core 6066ce0d and Package 348b98b2: 1006 bindings, 172 shared tests, 1525 Module tests, 116 schema parity tests, 37 Transfer tests and two Default-Core durable cases. Pending intents survived event cleanup and refused actions before and during Escrow. Only the exact claim with a ready arrival completed. Root verified the saved records and Wasm SHA-256 352943d1468b1e7f5464e9437a097bfdd293d6894f0b09d45404c52c543f88ab. Full Package lifecycle and real routing remain pending.
• Review at Core 27c74030 clears the bound-old-bot and same-Shard settled-retry corrections. Two remaining races require source changes before runtime: human arrivals must retain the original source partition and Realm revision, and the generic release reducer must atomically refuse a bot-owned arrival fence. Gateway cache checks alone cannot protect that release. The owner is correcting both and the in-memory Store retry contract.
• Both source reviews clear Core d153f4c1 and Package a5171eeb. Human and bot arrivals retain the exact source partition and Realm revision through settlement and release. The human release preserves Session Actor and Account Claim Gates. Supported instance AreaTrigger handling is now in source development; this revision has not yet passed runtime checks.
• Source review of Core 21154017 and Package c4714892 found four gaps before runtime: arrival reconciliation reset a deferred objective deadline and retry count; Legacy bypassed checkpoint normalization; an unclaimed Transfer Intent did not keep its instance lease occupied before Escrow; and malformed portal dimensions or landing coordinates lacked finite checks. The owner is correcting these paths and adding focused caller evidence. The earlier Escrow phase already protects instance occupancy.
• Core b9b50b2b closes the reviewed pre-Escrow lease and malformed geometry gaps in source. Further review of the proposed Legacy migration found a dead companion could bypass resurrection, a blocked approach could bypass Recovery, and a same-Shard crossing could still use direct placement. The owner is replacing that partial path with the shared decision and arrival operations before runtime acceptance.
• Offline extraction of the pinned client DBC through the actual importer confirms trigger78 at map0 (-11208.5, 1685.34, 25.7612), radius7. Exit119 is on map36 at (-14.3628, -393.38, 64.5605), radius6, only 8.183 yards from the imported entry landing. Exit121 is 631.694 yards away and is excluded from this short scenario. These are exact source values, not navigation or client acceptance. The exit floor remains unverified.
• Package d4df4bf removes the separate Legacy Transfer executor and shares Recovery, next-eligible times and checkpoint handling with the main runner. Follow-up source review found the crossing policy still waits or returns home when the leader is already in the same dungeon, and the stranded wait can postpone death recovery. These need correction before acceptance. Default-Core bindings generated successfully; focused compilation and tests are correcting stale schema-accounting expectations. No latest route runtime is claimed.
• Both source reviews now clear Package5587f2fa: a bot already beside its leader stays in the dungeon, and a dead Legacy bot reaches Resurrection before the stranded wait. Core0de208a7 corrects the Realm Store interface and its in-memory ordered location model, including pending refusal and atomic revision updates. The focused default-Core check is running; final Package route and restart acceptance remain pending.
• At Core 0de208a7 and Package 5587f2fa, all 1,532 Module tests and 172 shared tests passed, and the sessionless durable target compiled. Gateway compilation found four generated API call errors. Core 7460c2ac corrects the transfer index, borrowed roster keys and missing table trait; source review is clear and the focused Gateway gate is running. The acceptance map still marks the full entry/exit route, real restart boundaries and destination recovery callers as incomplete.
• Gateway compilation passed at Core 7460c2ac. Schema parity then found three stale field lists. Core b7d5c5a4 corrects those lists and the exact human arrival fixture identity; source review is clear. Separate callers now cover source cast cancellation past its deadline and party mirror failure with the retained Realm predecessor. Their destination-action and real-process acceptance portions remain open.
• The trigger 119 caller now records exact per-case source provenance, map 36 sphere dimensions, instance admission and destination coordinates at Core ce85e820 and Package a4432cb. Review corrected SATS action-kind comparisons before runtime. Separate lifecycle source at Core 713223a2 and Package e6f7024 covers actual Module replacement during a cast, process restart during an owned movement leg and provisioning, and a completed Quest reward across replacement. Movement and provisioning reviews are clear; the Quest caller is under final review. None of these new callers has run yet.
• The frozen default-Core Gateway gate is running at Core 81ddcb7b and Package bf23cf82. Review first corrected the route caller to match actual SATS Option, action-kind and empty-vector text. The separate four lifecycle callers are frozen at Core 57d097f7 and Package e6f7024 for the next private union run. Actual merged PB-009 composition and final CI remain required.
• The Gateway gate at Core 6edc3333 and Package bf23cf82 compiled and passed 38 of 42 Transfer tests. The saved-destination case exposed use of a stale caller plan before Realm settlement; Core 81bc52cf reads the existing Escrow first and passes source review. Three fixture corrections address the earlier exact phase refusal, admitted instance lease and mirror interruption boundary. A separate proposed abort matrix now names twelve bot phases; actual process evidence remains pending.
• The first four-case lifecycle union at Core 57d097f7 and Package e6f7024 failed before runtime: the stored Transfer purpose used unsupported named enum variants, and the Package read a trigger-id field absent from the Core route API. Reviewed Package c9d780e7 preserves the enum and field order through typed payloads and carries the original trigger id beside the route. Native compilation then passed in 43.50 seconds. Debug Wasm found one missing party-partition table import in a private fixture. Package 95fc0af9 corrects it and removes two unused reducer exports. The third build is running; both earlier failures are retained.
• At clean Core 57d097f7 and Package 95fc0af9, native compilation passed in 43.81 seconds and debug Wasm in 1 minute 40 seconds. Cast retention across Module replacement passed in 34.97 seconds, provisioning across process restart in 9.32 seconds, and completed Quest reward retention without duplication in 26.20 seconds. The orchestrator verified twelve runtime records and both actual replacement Wasms against the clean source pair. Base Wasm SHA-256 is 4466911f8665c7fbe701b20138a35f8d2adf8893780024e7f7bf1cc83e400afd.
• The movement case stopped before restart because its first decision created the return-home objective and deferred movement until the next eligible time. Reviewed Core 664254c1 waits through ordinary decisions until an actual movement leg exists, then retains the original restart and completion assertions. It does not alter the declared Core tick, positions or runner state. The movement-only rerun is queued.
• The corrected Gateway gate at Core 3fcba9db and Package bf23cf82 compiled and passed 41 of 43 Transfer tests. Two resumed cases exposed a test Store mismatch: it rejected an already settled exact crossing that the durable Realm operation accepts. Core 396de7f7 corrects this exact retry and has passed source review. Runtime remains pending; both failed records are retained.
• Movement restart passed in 35.33 seconds at clean Core 664254c1 and Package 95fc0af9 after native and debug Wasm builds passed. The orchestrator verified the changed process id, unchanged program and schedule, identical retained movement leg across restart, its completion at the old endpoint and a later ordinary leg from that endpoint. All four lifecycle behaviors now pass at the same base Wasm SHA-256 4466911f8665c7fbe701b20138a35f8d2adf8893780024e7f7bf1cc83e400afd. Final composed-source CI remains required.
• The Gateway gate passed both focused retry cases and all 43 Transfer tests at Core 396de7f7. One missing roster-revision parity entry then failed the schema suite. Reviewed Core aa58e27e adds its three fields and manifest entry; all 117 parity tests and the check of 1019 current bindings pass. These tests do not replace the prepared real Gateway process-abort cases.
• Source composition has started from the command Core and Package revisions. New callers cover twelve actual Gateway process-abort boundaries, arrival party-mirror refusal, destination Quest rebuilding and incompatible content, plus the four companion orders at the audited entrance. Reviews and correction work continue before the combined build. No new route or Gateway process runtime result is claimed.
• Review of the new real Gateway process caller found an unreachable final failure point after the ready-arrival early return, plus missing exact signal, pre-release gameplay and crossing-identity assertions. These findings are being corrected before its first runtime. Earlier 43/43 Gateway tests remain valid at their recorded seam; they do not prove the new process caller. A separate Transfer CI job generator passes source review and static YAML/shell checks, with actual merged pins and both CI runs still pending.
• Source review clears the selected-member correction at Package 2cbc21d3. Assist now follows its named member during ordinary movement, fallback movement, crossing and arrival. Core 50a0f62b also retries instance cleanup after a crash following arrival release. The new Gateway process caller has stronger crossing and gameplay assertions at d9df42db; remaining process identity and exact party-state evidence are being corrected before runtime.
• Source review clears the composed transfer cases at Core a0deca513 and Package 34a07570, with a narrow API exemption correction queued. The private Realm fixture now mirrors all four party members. Quest cases execute the production Transfer operation, cross through a real Gateway, then leave the destination Module party mirror before ordinary Quest reconciliation. The source and destination catalogues use different navigation revisions and targets. These cases have not compiled or run on the combined pair. They do not prove a Realm Leave command.
• The first combined builds exposed three integration defects: a missing group-function brace, Actor exports missing required macro aliases, and missing fields in the bounded command-roster read. The owner corrected them and regenerated the default Gateway bindings. At clean Core 8dc70084364cded63e8c8f5300d25e9013ae66a4 and Package abbc79bd45df8bc85331aafb6f7966102f48d341, 43 Gateway transfer tests and 120 schema parity tests pass. The target compiles with four real Gateway process callers and three shared utility tests, and all 1044 binding files match generation. Actual process scenarios and the full Package runtime remain pending. Earlier build failures are preserved.
• The full four-Package union compiles all four Module transfer test targets and builds the debug release Wasm at Core 72b3e76b and Package ba3007fb. The saved Wasm has SHA-256 229d14e967341d262adf51e572791cd8ab5e37e37a26ccf1359c9a08bbf0db83. Strict Clippy found remaining source issues; corrections and the named-member Assist arrival scenario are under review. The orchestrator is building the actual merged command predecessor for the populated upgrade case. No new combined process runtime result is claimed.
• The actual merged command predecessor now builds from clean Core e025ab7696f557ff10a88a1bb33710e3066aedc3 and Package 7858c7fdd6d0d4dd49b431419360e8ed22d69388. The full four-Package debug release Wasm is 8,641,035 bytes with SHA-256 81bedd04baaab8e1c1598c1418320669bd99257b676bc35a491b3a27dbfd974a. The orchestrator verified the complete build manifest, source identities and cleanup. This is the input to the populated upgrade case; it is not an upgrade runtime result.
• Review of Core 2d3c38c5 and Package 0bb8d562 found process-fixture facts that conflict with the real party Gates: different rules at the same roster revision, absent member locations, a missing destination holder and no destination mirror before fault injection. The injected cursor change also cannot cause the claimed mirror Refusal. The owner is correcting the fixtures while preserving production checks. Assist now has assertions for the selected Priest, exact Realm party state and actual Core movement; final source review is pending. These are source findings, not completed process runs.
• The corrected process fixtures pass root review and both independent reviews at Core 2ac6cb6853f05ed8e627d3db1a7f961bc2f69b5f and Package a7a8ea05e924a14f799dd93ec61f59f2a6262b78. Generic cases now stage valid Realm locations, a real destination holder and an accepted mirror before the checked-rule fault. Both World mirrors compare exact semantic membership. Assist preserves the selected foreground and movement leg before measuring actual Core movement toward the Priest. Its controlled decision pause is explicit. The required default-Core gate is running; no process runtime result is claimed.
• At clean Core 2ac6cb6853f05ed8e627d3db1a7f961bc2f69b5f and Package a7a8ea05e924a14f799dd93ec61f59f2a6262b78, the default-Core gate passes 43 Transfer tests, 120 schema checks and generation comparison for all 1044 Gateway binding files. The process target compiles with five scenarios and three utility tests. The orchestrator independently verified the logs and source identities, then opened Core PR 503. Full Package process runtime and required PR CI remain pending.
• Core PR 503 at 10c75fc6 passes formatting, Clippy, build, unit tests and the Wasm build. The durable job is still running. Automated review posted twelve findings; two lint findings are already resolved, and the remaining findings are under source review. A successful review status does not mean these findings are closed.
• The first final four-Package attempt at Core10c75fc6 and Packagea7a8ea05 stopped during the first Module compile because a private instance fixture call lacked its Package API exemption. No runtime case executed. Both cleanup phases passed, and the original failure logs are retained. Narrow fixture annotations are corrected before the next combined run.
• Required durable CI at Core10c75fc6 failed the existing Character cleanup scenario because its setup omitted the source Realm locator. The reviewed correction records and observes all three party members on their actual holder maps before Transfer. This changes the fixture only. The failed run is retained; the corrected scenario must pass before merge.
• At Core71761b83 and Package49b9bad6, all 46 Gateway Transfer tests, 85 Realm-core tests and 120 schema checks passed. Default generation produced 1045 binding files. The local durable and binding-comparison runs then failed during the Wasm build because the temporary filesystem quota was full. No runtime progress is attributed to that attempt. Completed temporary checkouts were removed with their branch and commit references retained.
• Coref26e4635 passes required CI formatting, Clippy, unit tests, build and Wasm checks. Its local durable cleanup scenario then reached actual Transfer and failed because the Gateway returned the old Known locator after observing the new Pending locator. Reviewed Core5d84e352 returns the exact observed Pending state. The corrected durable scenario is running; its result and the full sixteen Module and five Gateway process cases remain pending.
• The next durable cleanup attempt passed the repaired Realm begin step, then refused an unstaged instance. Core02aa4057 corrects the fixture to use its declared destination. That run reached arrival synchronization and exposed a production defect: destination import applied ordinary party-departure cleanup to an existing mirror, changing its leader at the same Realm roster revision. The failed attempts remain recorded.
• Core d20bfec9 detaches the destination party mirror before replacing stale Character rows. Both source reviews clear the change. Its focused import regression and adapter check pass, and the fresh debug release Wasm builds. Required CI formatting, Clippy, unit tests, build and Wasm checks pass. The durable scenario is next. All twelve PR review threads are resolved; runtime acceptance remains open.
• The durable cleanup scenario passes at Cored20bfec9 in 43.41 seconds using fresh Module Wasm SHA-256 ced4d26c647ba5401f5939b287f657e8d9a94e717a26b8feb996c396726d929d. This exercises both corrected production paths. A new immediate party-order assertion then failed at Core0d66e5af because the Gateway reader lacked the required membership-revision subscription. The failed run is retained.
• Core d35d6884 adds the configured subscription. All four subscription-policy tests and the durable cleanup scenario pass, including Realm join order on every World Shard immediately after arrival. The durable case took 42.96 seconds. The orchestrator verified its source identity, logs and retained Wasm. All 1045 bindings match generation; Module and binding source are unchanged by these Gateway corrections. The final four-Package build and runtime has started at clean Cored35d6884 and Package49b9bad6.
• Core d35d6884 passes four Coordinator query checks and the ordinary Character-gone Transfer scenario in 42.96 seconds, including exact Realm join order immediately on every World Shard after arrival. Its Module and bindings are unchanged from the verified debug Wasm and 1045-file binding comparison. CI then required a schema-parity entry for the added subscription. Reviewed Core be2d3294 adds all eight partition fields and independently checks its three enum tags. No production behavior changes in this correction.
• The combined attempt at Core d35d6884 and Package 7d33233a passed all five native test targets, then debug Wasm compilation failed with twelve fixture errors before runtime. Reviewed Package e9069bba corrects a table-trait import, uses exact eight-field partition snapshots instead of cloning a Core row, and types a group key as u64. Both source reviews preserve selected-Priest Assist assertions and exact party membership. The next combined run is pinned to Core be2d3294 and Package e9069bba, with sixteen Module cases, five Gateway process cases and three Gateway support tests. Neither failed build supplies runtime acceptance.
• At Core be2d3294 and Package e9069bba, all five native targets and the debug Wasm build passed. Actual runtime passed seven of eight Transfer cases, all four lifecycle cases, two of three order cases and all three Gateway support tests. The populated upgrade failed, and all five Gateway process cases failed in their shared Realm setup before process-crash assertions. The normalized-state failure is an incorrect progress-age expectation; the existing objective projection requires zero elapsed time in that pass. The shared Gateway setup updated the leader locator but omitted its partition projection. Target-local and upgrade failures are under diagnosis. All raw results are retained; Wasm bytes and source identities match after runtime, and owned-process cleanup passed.
• The three Module failures were stale assertions. Core a95abd18 now checks zero elapsed time for a newly retained objective, the exact cancellation stop spline for remote Target, and both rewarded prerequisite783 and active Quest7 across populated upgrade. Package 11df6778 stages the leader through its existing completed-crossing fixture operation, including the matching party partition. Source review clears all four corrections. A fresh combined run is executing at these exact revisions. Core be2d3294 has separately passed every required CI check, including the durable job; the corrected Core revision still needs its own checks.
• The corrected combined run at Core a95abd18 and Package 11df6778 passes all sixteen Module scenarios and all three Gateway support tests. All five Gateway process scenarios fail at later checks. The real Gateway instance-lease read has no subscription for game_instance, so a present lease appears absent and blocks Transfer. The other findings are fixture assumptions: a normalized source Runner remains until finish_transfer, ordinary Assist decisions append a runtime outcome after command acceptance, and every fresh Module seeds one weather metadata row. Corrections preserve actual lease, imported-content and action-fence requirements. Wasm SHA-256 130dc4f25d0e294a88a18fce2f65f7d544a7bad7ea9ea3b157c97cc78e0af174 matches after runtime; all source, capture and cleanup checks pass. Gateway process acceptance remains pending.
• Final source review clears Core4cb033ea and Packagee1731e33. The instance subscription audit accounts for all eight newly cached tables and finds no remaining missing query. The source Runner, accepted Assist history and exact weather-seed corrections preserve their gameplay and imported-content requirements. Two local compile attempts reached no tests after an omitted build setting exhausted temporary quota; their logs are retained and only owned build caches were removed. The corrected command now asserts the complete build configuration. All122 schema tests and four Coordinator query tests pass, and the combined runtime suite has started again on these final revisions.
• The third combined runtime at Core 4cb033ea and Package e1731e33 passes all sixteen Module scenarios and three Gateway support tests. All five Gateway process cases fail at later stages: Assist has no destination decision pass, the restart assertion treats an observation receipt as executable work, the mirror fault expects a member already detached by import, and both Quest setups attempt to exhaust Recovery before an active attempt exists. Evidence capture, unchanged Wasm bytes and all six owned build cleanup commands pass. No configured realm was changed.
• Core fd2c494f strengthens exact durable phase counts and requires the intent-bound TransferAccepted observation through source finish. Movement, Cast and melee work remain forbidden after Escrow. Three saved runtime phases match the corrected counts; later phases require a rerun. Core d7a27650 adds explicit real Runner passes while the Assist fixture delays the shared Core tick. The caller records this controlled decision driver and still requires actual body movement toward the selected Priest. Package b403248 changes only the mirror fault fixture to require the exact detached three-member party and its retained arrival identity. Independent correction review is in progress.
• The combined corrections pass independent Standards and Spec review at Core cd732f33b99a1cc84bedd827c665037eca156914 and Package 870435696794a27be46856c3a0d65889d6f87071. The Quest setup places the Character beside a real blocked target, requires an observed active Fight before exhausting Recovery, checks ten declared navigation cells and preserves setup evidence on timeout. Source and destination assertions derive retained objective age from its published timestamps; arrival must preserve the exact source objective and clocks before its first pass. The Assist correction and detached-mirror fault fixture also pass independent review. The next local run builds fresh combined Wasm and repeats the five Gateway process scenarios plus three support tests. These corrections have no new runtime result yet.
• The first focused Gateway build at Core cd732f33 and Package 87043569 compiled the test binary but stopped before runtime on an incorrect private Transfer accessor import in the mirror fixture. Package 44f1192 corrects the owning import path and records its narrow private API exemption. The failed build and cleanup evidence remain available.
• The focused rerun at Core cd732f33 and Package 44f1192 builds fresh combined Wasm, passes all three Gateway support tests and passes selected-member Assist through actual Core body movement. Four process cases reach later failures. Restart and mirror repair compare certified World locations with stale location fields in the Realm party projection. The Gateway correctly certifies those locations from current Realm Character locators and healthy holders. Both Quest cases pass blocked-target Recovery setup, then route staging conflicts with the accepted party roster revision after Leave and rejoin. These two causes have separate correction owners. Evidence capture, unchanged Wasm bytes and cleanup pass.
• Core 6fc104e5 and Package fc45fbea contain the independently reviewed corrections. World party assertions now require the exact four members, current settled Realm locators and matching serving World Characters. The Quest route fixture preserves accepted membership order, and its Leave and rejoin operations advance Realm and World through the same history. A separate Core CI failure exposed a fixture race after staging Transfer; the corrected test waits for the Coordinator to observe that location before starting its worker. The focused deletion case and five Gateway process cases are rerunning at these exact revisions. No new runtime pass is claimed yet.
• At Core 6fc104e5 and Package fc45fbea, the focused deletion case passes and all required Core CI checks pass. The fresh four-Package Gateway run passes three support tests and three of five process cases: selected-member Assist, all twelve committed restart stages and mirror repair before arrival release. Both Quest cases now reach the real source Transfer operation, then their snapshot requires a source body that portal entry correctly removes. A narrow capture correction is in progress. Wasm SHA-256 ea26113d2a2a7b79bf342f9c100f46b2eed5571bc7d61f339e63dc0d446a4d25 matches after runtime. Evidence capture, cleanup and source checks pass. The two Quest cases still require a successful rerun.
• Core 4e69794e and Package fc45fbea build successfully. Both focused Quest cases pass the source lifecycle and retained-purpose assertions, including one body before portal entry, no body afterward, the exact Character and Intent destination, unchanged objective and Quest, and all ten source navigation cells. Both then refuse destination catalog setup before Gateway crossing. The generic destination creates bots, which initializes a catalog header, twelve derived Quest rows, twelve objectives and three class seeds. The later private fixture incorrectly requires those rows to be absent. Separate owners are correcting its handling of the exact generated catalog and recording setup state before the operation. Capture, cleanup, source integrity and unchanged Wasm checks pass. The two Quest process cases remain unaccepted.
• Core 1f8becaf and Package 32fa1f4f build successfully. The private destination setup now validates and replaces only its exact unused generated catalog, with bounded reads and transactional rollback. Both Quest cases pass source retention, catalog setup, Gateway crossing and arrival retention. They then lose their retained Quest during destination decisions. The compatible case records MissingActualEndDestination for Quest7; the incompatible case records MissingStartDestination for Quest5261. Production admission searches map0 for the giver and ender even though the destination catalog and creatures are on map36 in instance5098078. A correction must preserve both map and instance and retain current-content checks. Both cases fail, zero of two, in 74.55 seconds. Wasm SHA-256 46d4cbb7d067ef10fd3337b5f673d389c3ee479806accf4ea6ae53ce435d6285 matches before and after runtime. Capture, cleanup and source integrity pass. No configured realm was changed.
• At Core 1f8becaf and Package 7a122678, both focused Gateway Quest cases pass in 65.44 seconds after a fresh four-Package build. The compatible case keeps Quest7, its objective identity and deadline, then rebuilds Recovery from destination target and navigation evidence. The incompatible case records TransferPurposeChanged, retains replacement Quest5261 and accepts it through the normal Quest operation. The correction validates current giver, source and ender evidence in the arrived Character partition and requires exact catalog matches. Root and independent source reviews pass. Wasm SHA-256 b5db2eeb79cc55fbddce528f5274654b3667465e755c27895fa92aeb1611dbe2 matches before and after runtime. All capture, cleanup and source checks pass. Core 1f8becaf also passes every required CI check and has no unresolved review threads.
• Four existing map0 catalog regressions pass at Core 1f8becaf and Package 642e34d9 after a fresh four-Package build: creature and provided-item paths, GameObject loot and simple use, stored destinations without live targets, and retained Quest work across combat and changed content. Every exact case runs and passes. Capture, cleanup and source integrity all pass. The Wasm bytes match the successful two-Quest transfer build. The Package README now defines deletion and Transfer for all ten Character-owned and seven Shard-owned production tables. Independent review confirms the table and crossing policy match source. The final copy distinguishes retaining a replacement objective from executing it after the recorded arrival Refusal.
• The final instance entry and exit caller passes both source reviews at Core 93efc7a1 and Package b0b0f69c. It authenticates Follow before trigger78 entry, drives the companion through trigger119 exit with the real Gateway, and requires the exact retained order, objective identity and deadline. It checks two generation advances, settled Realm locations, matching World party mirrors, old source removal, arrival release and actual Core movement toward the returned leader. Review corrected the objective witness and asynchronous movement assertions before the frozen run. The focused check uses freshly built full four-Package Wasm and one exact ignored case. Final Core checks and runtime remain pending. The complete Gateway target now contains six process cases and three ordinary support tests.
• The first exact exit run at Core 93efc7a1 and Package b0b0f69c compiles the Gateway target and fresh four-Package Wasm, then fails before entry in 10.14 seconds. The fixture submits an authenticated Follow order but stages the portal with a tokenless leader Actor, so the normal instance admission Gate returns STALE_WORLD_SESSION. The correction carries the same caller Actor through that private setup while preserving existing sessionless callers and all production ownership Gates. The retained run reports zero of one passed, unchanged Wasm SHA-256 ddc3cfd94da7b763d5edb47ea8cd6c0625653a4740b34b95e217f9d83c14154a, successful evidence capture and cleanup, and unchanged source inputs.
• The corrected authenticated setup at Core 1debba98 and Package df62213c builds and completes the actual entry transfer with retained Follow, objective and deadline. The exit setup then fails because it expects membership revision5098005 from the separate Quest Leave/rejoin scenario. This caller performs no Leave or rejoin. Its saved Realm and both World records retain membership5098002 while World-local row IDs change to5. The next Package correction checks that exact retained history and passes both source reviews. The failed case reports zero of one passed in 20.30 seconds. Wasm SHA-256 78ff297ecbe348d00ab5b094b7681ef17c6fa80fd6c8e2700b8aa337e6f671c1 is unchanged; capture, cleanup and source checks pass. Core remains 1debba98 and the corrected composed Package is 2a8b3223. Actual exit and final canonical CI remain required.
• At Core 1debba98 and Package 2a8b3223, the corrected route reaches the exit Transfer. The assertion incorrectly expects one action observation across all kinds. The saved rows contain one completed cast and the exact accepted Transfer observation, with no live source body, movement, pending cast or melee work. The correction selects the Transfer kind and preserves the source-work checks. This attempt reports zero of one passed in 26.82 seconds; capture, cleanup and source checks pass.
• At Core 6bfb7f5a and Package 2a8b3223, real entry and exit complete. The first arrival snapshot retains Follow toward the exact leader, objective1, generation3 and the original deadline. Its Core spline starts at the imported exit landing and the body progresses toward the leader. An unnecessary later Runner pass completes Follow and removes the spline, so the old polling assertion fails in 51.62 seconds. Core 36db241d retains the already-saved arrival witness, verifies its destination and later physical progress, and permits an already-completed spline. Independent Standards and Spec review is clear. The fresh exact run is in progress; no final pass is claimed. Both attempts preserve Wasm SHA-256 9c824bc9fa223e138c54131450ae4751341a81413a50b8af5ca8058881048e89 and complete cleanup.
• The fifth run at Core36db241d completes entry, exit and Follow movement but fails because it expects the recurring 0.5-second scheduler row to change. That interval row is constant. The correction ties the exact Move observation to the saved foreground start and verifies the same observation through actual body progress. Both reviews clear it. The sixth fresh run at Core73617467 and Package2a8b3223 passes the full case in36.81 seconds, including the remaining order, deadline, party and location assertions. Capture, cleanup, source checks and all process statuses pass. Pre/post Wasm bytes match SHA-256 9c824bc9fa223e138c54131450ae4751341a81413a50b8af5ca8058881048e89. Configured realm processes are unchanged by the run. Final Core CI and canonical paired Package tests remain before merge.
• Core PR503 merged normally at 8183c8a30a080a0d9a789db40256824f78b781e2 after all required checks passed and all twelve review findings closed. Package PR34 pins that actual merge. Its Module job passed 1650 tests and failed two Architecture Tests in debug fixtures. Reviewed Package f5c25e8 replaces nine raw Character reads with the escrow-aware helper and defines the Assist source evidence lifetime. Focused union checks are running before the correction is pushed; canonical Transfer and regression jobs remain required.
• At Core8183c8a3 and Packagee6131108, all 1652 Module tests pass, including both Architecture Tests. Sources are exact and clean before and after; command, capture and final status are zero. Canonical CI separately found outdated party-command setup and an entry test counting an old Move observation as a Transfer receipt. Both Core test corrections pass source review. Package review also confirmed retained quest positions could cross instance boundaries and Legacy arrival could retain its checkpoint indefinitely; those corrections and focused runtime remain open.
• Core PR504 is open at 9dda6539 with the reviewed party setup, Transfer observation check and held-Quest partition regression. A fresh four-Package run is executing six party-command cases, the full entry/exit route and the new Quest case against Package c1eba84f. The separate arrival recovery correction is still in implementation. No pass is claimed for this new run.
• The focused 9dda6539/c1eba84f attempt stopped during compilation at the temporary disk quota; no tests ran. Its failed status and cleanup evidence are retained. Sharing identical saved Wasm files recovered 607 MB while preserving paths and checksums. Independent review cleared the arrival correction after adding live-member checks and saving staged evidence before assertions. Core PR504 now contains five arrival cases at 1f1140ea. Combined Package 78bb0909 includes the architecture, Quest position and arrival fixes. A 13-case local run is being prepared; no new runtime pass is claimed.
• The combined 13-case run at Core 1f1140ea and Package 78bb0909 completed with eight passes and five failures. Full instance entry/exit, held Quest partition checks, two arrival cases and four party-command cases passed. The failures came from a dead-member staging call, two empty-vector assertions and absent authoritative Realm locations. Those fixtures and assertions are corrected. Capture, cleanup and source verification passed.
• Core PR504 now holds the corrections at 601cc987. CI at its previous head caught one evidence JSON macro above the Rust recursion limit; the field now uses a separate assignment with the same recorded shape. Package 81573fda is frozen for the reduced eight-case retry. No result is claimed for that retry before it finishes.
• The reduced run at Core601cc987 and Package81573fda passed all three affected Gateway command cases and four arrival Recovery cases. The remaining assertion incorrectly prohibited ordinary Follow spatial arrival while Transfer Recovery remained pending. Independent diagnosis confirmed the checkpoint, objective and compact budget were retained. The corrected exact-state case passed at Core2dec010b in 4.61 seconds. All five Recovery behaviors now have passing local evidence; fresh Wasm bytes match across the last two runs. Capture, cleanup and source checks pass. Core PR504 has no unresolved review threads; its last CI job remains pending.
• Core PR504 merged at 855d96c858331d604da9a262510daa89302f53d2 after all CI checks passed. Its merge tree matches the locally tested 2dec010b tree. Package PR34 now pins that actual merge at cab699de and adds all five Recovery regressions to its twenty-one Module Transfer cases. The four Package review threads were answered with the pushed corrections and resolved. Final canonical Package CI remains in progress.
• At Package cab699de and merged Core855d96c8, canonical Transfer CI passed all six real Gateway process scenarios and all twenty-one Module cases: eight Transfer, four lifecycle, three order, five Recovery and one populated upgrade. The broader Package regression job is still running. CI run 34459875814, job 102814924803 retains the exact output.
• Both broader CI runs completed with failures at Packagecab699de: three shared assertion or setup failures and one timing-dependent casting-position failure. The corrected four-case run at Core7123b249 and Package2ae6e30f passes every case. Native compilation, the fresh four-Package Wasm build, capture, cleanup, source and Wasm identity checks pass. The casting fixture now places the body and spawn atomically and retains all eight measured casting-position legs.
• Core PR506 merged normally at c147b0e4a0a8734287a4f7452397a069d5aeefb8 after every required check passed. Its tree matches the tested Core7123b249 tree. Package PR34 headffb14b42 contains the same tested playerbots content, with both current CI jobs pinned to that actual merge. Independent source review passes; fresh canonical Package CI is running and the ticket remains incomplete.
• Final Packageffb14b42 Transfer CI passes all twenty-one Module cases and six real Gateway scenarios at merged Corec147b0e4. The exact result groups are 8, 4, 3, 5 and 1 Module cases, plus 6 Gateway cases. Canonical run34467711841, job102840112590 retains the output; log SHA256a7e56883cecd730b573d9dbbb0085bafabbd53b3232d2307ebe51b443dfc5b31. Broader regression CI and the normal Package merge remain pending.
• Final Packageffb14b42 passes every required check in both CI runs. Canonical run34467711841 passes all146 broader cases across12 groups, plus21 Module Transfer cases and6 real Gateway scenarios. The broader log SHA256 is4a1f99a2e94aa441e77ce18652adfe2376cca57c0c8e3184285b9ddb3a9cdf9b. All four review threads are resolved. Package PR34 merged normally at2453f0125ec29b26105234c8e73891cbb53d8a33 on2026-09-10 at11:41:59UTC. Its tree exactly matches the tested head. All eight acceptance criteria now have reviewed evidence.

Current notes

• Source preparation has begun in isolated worktrees. It covers durable Transfer Intent claims, restart discovery and party repair before arrival release. Final composition, runtime checks and acceptance wait for merged recovery and commands.
• The first foundation is frozen at Core 63ff3f5b and Package 348b98b2 for independent review. It adds durable claims, bounded restart discovery, and exact source intent identity on the destination arrival fence at import. Root review caught a stale-worker race during design and a serialized-value assertion defect before runtime. Generation, compilation, the real restart matrix and full semantic transport remain pending.
• Default-Core generation passed for all 1006 binding files. Module checks passed 1525 tests, schema parity passed 116 tests, and Transfer checks passed 37 tests. Narrow compilation corrections and removal of unused Store methods are frozen at Core 6066ce0d. Two new durable consent cases are next. Full retained-purpose transport, real instance routing, restart composition and final merged dependencies remain required.
• Remote party location review found delayed snapshots could restore an obsolete partition. The next slice adds a Realm-owned locator revision and separate unknown, in-transit and known states, plus membership removal history. It must prove stale delivery, temporary read recovery and arrival before release without comparing clocks from different Shards.
• Independent reviews found four ordering defects in the party-location source slice: an old roster could remove a newer member, old leader and loot settings lacked an authoritative roster revision, a paused old Transfer worker could overwrite the current location, and a holder check could certify stale presence during Transfer. The owner is correcting Realm-owned ordering and exact crossing settlement before generating bindings or running this slice.
• The corrected source passes both reviews for full roster ordering and healthy holder certification. Three Transfer retry defects remain: an old bound intent can mark a newer Realm revision pending before refusal, generic recovery can settle a newer crossing to the same destination, and same-Shard retry cannot resume settlement after the ready write fails. Both reviewers require those corrections before binding generation.
• The orchestrator owns the Module replacement and persistent restart callers in isolated worktrees. The first cast replacement caller passes source review at Core 25c60f20. It appends an inert Wasm custom section, records both byte identities and requires the durable program hash to change before checking exact pending-cast retention and one completion. The movement restart caller at Core d680e989 and Package e6f7024 uses a declared future Core tick, preserves the real owned spline across PID replacement and waits for that stored tick to complete it. Source review and runtime of the movement caller remain pending.
• The combined order caller now passes source review at Core 56901c00 and Package 392bdb5. It binds exact order definitions, remote member partitions and the named enemy to the retained movement leg. It proves Target cancellation before the declared movement tick can finish that leg. The production composition will carry the selected Follow or Assist member through objective, checkpoint and arrival checks, and suppress automatic crossing for Stay and Target. Combined runtime remains required.
• The final Gateway target contains five ignored process scenarios and three shared utility tests. The four Module targets contain fifteen ignored scenarios. Acceptance will report these counts separately. The earlier seven-test compilation comprised four process callers and three utility tests; the added Assist case raises the process count to five.
• The combined review corrections require healthy Realm cache reads, preserve configured Realm authority, release the exact local arrival fence, retain the local instance population, and allow an exact portal intent replay only with current authority and consent. Overflow logs now report state transitions. The new replay scenario raises the Module Transfer suite to sixteen cases; five real Gateway process scenarios and three utility tests remain separate. Default binding regeneration and actual runtime are pending.
• The orchestrator retained strict repair of every connected World party mirror before arrival release. A proposed rollback cancellation path was declined after tracing production callers: cross-Shard escrows wait for import evidence, and the same-Shard bot path creates no escrow. Both review dispositions are recorded on Core PR 503.
• A separate source correction is under review for World party reads after arrival. It orders members by the retained Realm membership revision, so a newly allocated local mirror row cannot change join order. The existing durable caller checks this before Character deletion can hide the difference.
• The orchestrator owns Core PR503 and the final composition. The two Quest process cases pass at Core 1f8becaf and Package 7a122678. Normal Core merge, canonical Package CI and the Package merge remain required.
• The final acceptance audit found that the audited trigger119 exit case stops at its admitted Transfer Intent. It does not yet prove arrival outside the instance through the Gateway. A separate caller is being added to retain an authenticated order through that actual exit. PB010 remains open until this criterion passes.
• The final Transfer-specific CI remains green at Core855d96c8 and Packagecab699de. Broader PR CI fails three cases: Stay after Recovery, safe position after partition change, and GameObject approach after restoration. The duplicate push run also fails one casting-position progress assertion. Exact logs and uploaded evidence are retained. Root is correcting the three confirmed fixture expectations; the fourth is under independent timing diagnosis. No merge occurs before corrected CI passes.

Run the complete automated release suites on the pinned core, Package, imported content, and geometry. Prove seeded autonomous work, the fixed human-led companion route, restart and Transfer behavior, bounded recovery, scheduler fairness, and measured 10, 25, and 100 bot load.

Acceptance criteria

• Run ten seeds per starter class through the supported Northshire and Elwynn catalog. All ten runs per class must finish each supported scenario without Operator help: nine of ten is below the 95 percent requirement. Every failed run must record an explained outcome and fails the release gate.
• Autonomous acceptance includes kill, collect, talk-only chains, supported simple GameObject use, travel beyond the former home leash, turn-in, death, combat interruption, full bags, absent targets, an unreachable giver, and unsupported objective admission.
• One human-controlled Character and four bots complete the fixed repeated-pull route with explicit roles, follow and stay, assist and designated target, wounded allies and missing buffs between fights, loss of line of sight, controlled enemies, death, regrouping, and one instance entry and exit.
• The companion fixtures produce no unrequested pull, unexplained role abandonment, repeated cast start, or attack against an obsolete target.
• Every supported action covers success, waiting, Refusal, cancellation, and expiry through the highest suitable behavior seam. An unreachable outcome requires a reviewed owning contract and an executed check of the policy that excludes it.
• Restart and Transfer acceptance shows no duplicate Character, equipment, supply grant, quest reward, or controller and no stale action at the old location.
• Named disposable load runs execute 10, 25, and 100 bots and record decision time, movement time, scheduler lag, route expansions, row reads, row writes, and slowest due bot.
• The load report sets a measured deployment capacity, records fixture resources and revisions, and accounts for every deferred due bot without silent loss.
• All relevant repository checks and both specification and standards reviews pass at the pinned revisions.

Verified criteria: 9/9.

Run ten seeds per starter class through the supported Northshire and Elwynn catalog. All ten runs per class must finish each supported scenario without Operator help: nine of ten is below the 95 percent requirement. Every failed run must record an explained outcome and fails the release gate.

Reviewed by Orchestrator, with independent source and final artifact review.

• Canonical run 34569550383 passed all ten declared journeys per Warrior, Priest and Mage. Expected and actual case lists match, and all four required snapshot boundaries exist for every journey. Final evidence review SHA-256 c326b3f64b5f021945ae58e340572b6f271e1099e85b18e5e400357ad9a12361.

Autonomous acceptance includes kill, collect, talk-only chains, supported simple GameObject use, travel beyond the former home leash, turn-in, death, combat interruption, full bags, absent targets, an unreachable giver, and unsupported objective admission.

Reviewed by Orchestrator, with independent source and final artifact review.

• The same thirty journeys passed the supported Quest and fault scenarios with every required stage captured. The additional ordinary Warrior level-gap case passed in 421.02 seconds. This proves the declared fixture catalog; imported gameplay remains a separate observation. Final review c326b3f64b5f021945ae58e340572b6f271e1099e85b18e5e400357ad9a12361.

One human-controlled Character and four bots complete the fixed repeated-pull route with explicit roles, follow and stay, assist and designated target, wounded allies and missing buffs between fights, loss of line of sight, controlled enemies, death, regrouping, and one instance entry and exit.

Reviewed by Orchestrator, with independent source and final artifact review.

• Current composition job 103168568093 passed the fixed human-and-four-companion route and restart, Transfer and lost-acknowledgement scenario. Archive SHA-256 0be8b083b30e791a9cc7c85bd4a5f465a0bac2219eebc491adcb9af77d045974. Independent final review c326b3f64b5f021945ae58e340572b6f271e1099e85b18e5e400357ad9a12361.

The companion fixtures produce no unrequested pull, unexplained role abandonment, repeated cast start, or attack against an obsolete target.

Reviewed by Orchestrator, with independent source and final artifact review.

• Both current companion cases passed. The controlled-target snapshots retain exact casts and physical receipts through settling, then stable full receipts, health 832, one control Aura and no owned combat Handle. Held b89d4d24551bb71d965b233dd83d4ca18b65689cf0202a93c1883b9db8a32a20; settled 2f272c7c034b6d62dd2ae9ba71b381eba3807ac54d4d565a779529b6d9ec7a39; stable 90016c53ee9e4f5b2261a789f81a754f4151d9e0bd80962480b531b432d010a4.

Every supported action covers success, waiting, Refusal, cancellation, and expiry through the highest suitable behavior seam. An unreachable outcome requires a reviewed owning contract and an executed check of the policy that excludes it.

Reviewed by Orchestrator, with independent source and final artifact review.

• Current composition executed all 44 unique lifecycle cases and both companion cases with zero command failures. Its checker passed 15 actions, 75 outcome cells, two guards and two reviewed unreachable contracts. Job log SHA-256 da7db1e83a2b244021cbc2f3418eaac06fc33688f1800aabe3a1d753bd32bfa2; final review c326b3f64b5f021945ae58e340572b6f271e1099e85b18e5e400357ad9a12361.

Restart and Transfer acceptance shows no duplicate Character, equipment, supply grant, quest reward, or controller and no stale action at the old location.

Reviewed by Orchestrator, with independent source and final artifact review.

• Current Transfer job 103168568087 passed all 27 exact cases, and composition passed restart, Transfer and lost acknowledgement. Durable passed all 151 cases, including provisioning and recovery. Transfer archive 8d1f69bbde5ffeaa354878624e724043806ff922a42652970d584e16cc6e0e31. The prior reviewed run also exercised the authoritative-four mirror repair input.

Named disposable load runs execute 10, 25, and 100 bots and record decision time, movement time, scheduler lag, route expansions, row reads, row writes, and slowest due bot.

Reviewed by Codex orchestrator and independent evidence review.

• Persistent synthetic movement runs passed at 10 bots for 60 seconds, 25 bots for 90 seconds, and 100 bots for 300 seconds. Every bot was processed, progressed and crossed the declared navigation wall on a verified path. The runs retained decision timing, Core movement duration, scheduler lag, route expansions, row reads and writes, deferred work and the slowest due bot.
• The 100-bot run recorded 9,568 decisions and 9,536 completed movement legs, with maximum scheduler lag 2.513268 seconds. All declared lag limits passed and maximum Standalone RSS stayed below 538 MB. Complete capture, source, build and process cleanup checks passed.
• Tested Core d0ab53864fca2b42b17429624c8cc496882ef19c and Collection 175c3ffc91956ddaac4c921e0ec66d0da65d0d71. Module Wasm SHA-256 f4601db44f83e8894664e9b9cc55a0652264564eec3f9238380046a7fe54988b. Core PR #505 merged with the tested tree: https://github.com/LyraCoreProject/LyraCore/pull/505
• Aggregate result SHA-256 769f8f95d244d6236f8baccba7e0ef31c1c09635e4b3cc2f2151f848499ae5b1. Independent evidence review SHA-256 a8b8c881e2d38f905c67a7b6cb3c9e1045ff68baa423e0be817802707801f5bb.
• This accepts the declared movement workload. Criterion 8 remains open until the imported-world hour establishes the supported gameplay capacity.

All relevant repository checks and both specification and standards reviews pass at the pinned revisions.

Reviewed by Orchestrator, with independent source and final artifact review.

• All nine canonical Package jobs passed at cb74cbd1 and Core 2347fca0. Both required module contexts passed after the cancelled duplicate module job was repeated; rerun review SHA-256 ff1b7c1f973a416e4fbc060816822efa204635c55e0eafab46d2787a184eb7c5. All six PR review threads were resolved. Core #512 merged as 20524be6 with tested-head ancestry and exact tree dfd33f4d. Package #35 rebase-merged as a77013d1 with exact tested tree 5a08f3f8 and playerbots content d1544374. Source and final artifact Standards and Spec reviews are clear.

The load report sets a measured deployment capacity, records fixture resources and revisions, and accounts for every deferred due bot without silent loss.

Reviewed by Orchestrator and independent raw capacity review.

• Measured capacity is 25 bots for the pinned persistent imported-world fixture, with eight CPU affinity entries and a one-GiB page pool. Peak Standalone RSS was 887,173,120 bytes. This does not establish a higher bot count or capacity for the configured Realm.
• The full hour has 7,179 passes and 90,583 exact timing joins. Every bot was processed and progressed. All 92 deferred observations are accounted for. Scheduling delay p95 was 229,621 microseconds and maximum 506,693, below the three-second bound. No Reducer transaction rolled back.
• Decision host elapsed time had median 0.702 ms, p95 25.577 ms and maximum 75.207 ms. The Reducer writer-time estimate was 16.78 percent, not exact occupancy or OS CPU. Queue p95 is an upper bound of 0.5 ms. The raw audit includes all row counters and histogram bounds.
• Independent capacity audit SHA-256 71bdeb9dd28b713c687415338eb1815715c5157d9019087362dea0d90e881efb; root acceptance e681419b347641418f4194c55f756ded100449405aa3c6a2d41a5d0b856deaa9. Core #519 and Collection #36 are merged with trees equal to their tested heads.

Execution model: gpt-6-astra, xhigh effort. Repositories: LyraCore and Official Package Collection acceptance suites, Headless Client protocol and world-transition scenarios, Disposable imported-world load fixture and measurement report, wire-harness.

Recorded evidence

• Both source reviews clear the first three whole-journey callers at Core 4f826d7b. Native compilation passed in 43.64 seconds, then Wasm compilation found a missing Package lifecycle export. The narrow correction is Package c1b6e976; its new clean attempt is running. No whole journey has passed yet. All failed build evidence remains saved.
• The first three whole journeys passed at clean Core 4f826d7b19a7695e284c65446c80cddebed13149 and Package c1b6e97640876bedbddf51e3d90298f414df5de3. Warrior took 161.13 seconds, Priest 156.22 and Mage 133.89. Each completed all 13 seeded quests with exactly one reward, ten observed named-target deaths, real collection and travel beyond the former home leash. Root verified twelve source records, fifteen runtime files and the preserved Wasm, SHA-256 d7dce338e2bbc1274b7c4c20dea82f8ae27bd1f6619f42465936961b2a9422a4. All fixture commands stopped at ordinary Cohort activation.
• All thirty autonomous seed declarations now match the published class, role, position offset, due phase and source rotation plan. The first three journeys are the only completed whole-journey runs so far. The new scheduled combat-fault slice needs corrections from source review before it can count as acceptance.
• Headless Client #14 merged after independent review and Linux, macOS and coverage CI passed at a9e407786fe971ca0af2033ce69b9e5cbe9a294c. A second private check observed a real row update as one deleted row and one inserted row. The confirmed subscription retained the exact old and new named JSON rows. The first immediate metric scrape was stale; that attempt remains saved. The new load observer has four passing native stream tests. Full load and timing acceptance remains pending.
• The new Headless Client load observer at 4ec278f passes thirteen native behavior tests and Clippy. It verifies due order, accounts for every deferred bot, permits recorded damage-triggered scheduling advances, measures real progress and movement duration, and avoids recounting retained route expansions. It has not run a load fixture yet. Debug decision timing at Package 33999b5 passed independent source review; compilation and actual emitted timing logs remain pending.
• The thirty-seed combat-fault corrections and full-bag fault are source-reviewed at Core a9985073 and Package 357f7d0. The full-bag scenario retains the exact refused loot source, Quest identity, credit and retry deadline, then frees one filler slot for an ordinary retry. Actual retry timestamps are being added. These new fault scenarios have not yet passed runtime acceptance.
• Debug decision timing at Package 33999b5 compiled against Core a880bd29. A private off/on/off run emitted exactly one enabled sample and no disabled samples. Its actual CLI JSON names the Character, generation and runner timestamp. A second private run captured the confirmed subscription format. Both fixtures were removed after evidence capture. Observer 2f3b177 passed 16 tests and Clippy after corrections to progress boundaries, movement identity and route sample accounting; independent re-review remains pending. No load count is accepted yet.
• Acceptance preparation exposed a production quest-selection defect: reconciliation repeatedly selected a deferred destination. Package c2664ad now skips exact live deferred destinations without changing accepted Core quests. Root source review cleared the policy. The new unreachable-giver journey must still prove useful alternate progress during deferral, expiry and ordinary resumption. The full 30-journey matrix remains pending.
• A preliminary 60-second private movement run at Core a880bd29, Package c8d8633 and observer a703feb captured 595 decisions across ten bots. Every bot made progress; maximum scheduler lag was 4.233 ms and decision p95 was 944.269 microseconds. The transaction counter reported zero rollbacks. Navigation was disabled in this declared baseline, so route work was zero. The fixture was removed after evidence capture. This run does not establish imported or companion capacity.
• The first load run exposed a movement-duration measurement gap: runner arrival means the final destination, while Core movement uses short steps. Observer 94bcc079 now joins exact Core splines and entity endpoints to their runner identities. It passed 25 tests, Clippy and independent source review. A new ten-bot run with navigation enabled is in progress; larger counts and final dependency revisions remain pending.
• The unreachable-giver journey now requires an actual alternative Quest40 acceptance during its original 30-second deferral and Quest783 resumption after expiry. Prerequisite and unsupported-capability fault stages have passed source review after two evidence-state fixes. Ordinary combat across imported quest level gaps is in implementation. The full 30-journey matrix has not run yet.
• The second ten-bot run enabled navigation and captured 595 decisions and 585 completed Core movement legs in 60 seconds. Every bot progressed. Maximum scheduler lag was 6.175 ms, decision p95 was 976.35 microseconds and movement-leg p95 was 1.005412 seconds. The transaction counters reported no rollbacks. Route expansions remained zero because the flat route used the clear-path shortcut. The private fixture was removed. Package 83549b3 adds a declared short navigation wall for the next 10, 25 and 100 bot measurements; it has not yet run.
• Ordinary grinding source at Package 3da30213 selects suitable enemies when no supported quest is currently eligible. It preserves a suitable target across the eight-second selection rotation, respects Recovery and crowd control, and bounds candidate reads. Root source review is clear. Actual XP, level changes and admission at the unchanged imported quest level remain unproved until runtime.
• Headless Client PR15 merged normally as 114ac5af after Linux, macOS and coverage CI passed at reviewed head 576c719. CI ran all 81 generic tests, all 25 observer tests, Clippy and both release builds. Both source reviews are clear. The separate observer retains actual Core transactions, decision timings, movement identities and metric-window limits.
• The 60-second obstacle run passed at clean Core a880bd29, Package 83549b3 and observer 576c719. Before activation it verified navigation and all 134 blocked sub-cells. It captured 596 decisions, 586 completed movement legs and 1,159,168 route expansions. Every bot progressed, and an independent audit of saved Core positions found all ten paths passed around the actual blocked cells. Maximum search work was 4096 expansions, decision p95 was 15.313 ms and maximum scheduler lag was 4.269 ms. Counters reported zero rollbacks, 74,373 scanned rows, 4329 inserted rows and 4309 deleted rows. The private fixture was removed. Wasm SHA-256 5d2eb6a03c04b9e8032df9bd9d1af73a8e3565320e1fb71b8844746934901b01.
• The source-derived level boundary is frozen at Core ba16f507 and Package 84112598. It starts at declared level 5 with zero XP and synthetic enemies, keeps Quest40 at its actual minimum level 7, and requires ordinary kill XP, a live fight across target rotation and real acceptance at level 7. It explicitly has no imported-content identity. The first native attempt stopped at a missing inline annotation on a debug damage call; that source-only correction is clear and the new build is running. It is separate from the pending imported-world proof.
• The third level-boundary attempt at Core ba16f507 and Package ca7e037 passed native compilation and release Wasm, then retained four failed runtime records. A synthetic enemy with zero configured damage used Core fallback damage and killed the bot, which resurrected normally. Three journey cases stopped at an obsolete setup assertion that expected no prerequisite Refusal even though setup now records one. Core e892fd96 and Package f4c715fc correct the fixtures and are running again. No production behavior failure has been established by these four attempts. Imported-world acceptance remains pending.
• The fourth level-boundary run at Core e892fd96 and Package f4c715fc passed the live combat-continuity observation, then timed out at level5 with 1016XP after 11 recorded kills and resurrection. Diagnosis found a production decision defect: grinding excluded quest starters but selected quest ender952, which killed the bot. The correction must exclude any bounded quest relation and pass the unchanged enemy fixture. The kill evidence will also record every victim so an undeclared kill cannot be hidden inside the next XP observation.
• The representative Warrior journey exercised unreachable-giver and admission faults before the lethal interruption. It retained Quest7 and objective17 through resurrection, then timed out while walking back from a default graveyard about 6600 yards away. The synthetic fixture will declare a nearby Core graveyard before activation and prove ordinary resurrection and resumption there. This is a fixture geography correction; no post-start relocation is permitted.
• Root source review clears the leveling correction at Core e4612394 and Package 2ba40e3. Grinding now excludes any quest relation with one indexed presence read. The evidence records every bot kill before checking allowed victims. The synthetic journey places graveyard105 at its declared start before activation and records ordinary resurrection there. The corrected runtime is queued; the prior Warrior and level-boundary failures, interrupted Priest and unrun Mage remain recorded.
• The compressed observer passed all 29 local tests, Clippy and release build at e0867643. Headless Client PR16 then passed Linux, macOS and coverage CI and merged normally at b57904f34d367f6c96de07dbedf35182af5f5aa0 with no unresolved review threads. Both platforms ran 81 generic and 29 observer tests. The retained observer binary SHA-256 is 22aaec22dd51d233ae9bd06f8eac3ede10c55f237d212207ad0f6830798bcf63. No new load run is claimed.
• The observer is released as v0.1.0-alpha.12 at verified commit b57904f3 after its merged revision also passed Linux, macOS and coverage CI. The one-line Core pin is prepared for the final acceptance PR. A separate reviewed driver will measure synthetic wall movement on private persistent storage, for 60 seconds at 10 and 25 bots and 180 seconds at 100 bots. It verifies declared wall cells before activation and makes no SQL or fixture calls during measurement. These new runs have not started.
• At Core e4612394 and Package 2ba40e3, native compilation passed in 43.19 seconds. Debug Wasm compilation stopped because the fixture cloned a Core graveyard type without Clone. No journey ran. Package 51fb985d retains the graveyard scalars before consuming the row; source review clears this correction and the next focused run is queued. Failed build inputs and logs remain saved.
• The persistent movement driver and independent path verifier pass source review after requiring one west-to-east crossing per bot, refusing reverse crossings, checking retained executable and build hashes, and verifying the private persistent data directory was removed. They retain dynamic pass totals and complete gzip evidence. The final 10, 25 and 100 bot runs have not started.
• The corrected level and fault run at Core e4612394 and Package 51fb985d has compiled native and debug Wasm and entered the level-gap runtime. No completed journey result is recorded yet.
• At clean Core e4612394 and Package 51fb985d, the level-gap case passed in 410.72 seconds through ordinary kill XP while retaining Quest 40 at minimum level 7. Warrior and Priest then completed all thirteen quests, but their acceptance assertions failed: each damage event retained its own objective identity, while the test compared identities across an intervening decision. The missing-target fault also recorded removal but no required wait or resume observation within its declared two-second window. The Mage case is still running. These are incomplete fault proofs, and the failed records remain saved.
• The Mage also completed all thirteen quests and failed the same cross-stage identity assertion. Source and all three records establish an Available-to-Held Quest 7 transition between the two hits. The correction will keep exact retention checks around each hit and through resurrection. The Quest 18 source respawned after two seconds while the bot still had about 160 yards to approach it. The next fixture declares a fixed sixty-second absence before activation. Separately, the orchestrator audited all 68 level-gap kills: every adjacent XP observation matches, 6468 XP is accounted for, and Quest 40 is accepted at level 7 with 68 XP remaining. No imported-world claim is made.
• The corrected three-class fault run is source-clear at Core c8349f9e and Package 6e3ea9ec. It records the actual Quest 7 giver-to-held-target transition, exact identity and credit around each damage event, and the retained held destination through resurrection. The missing-target fixture declares sixty seconds before activation and records the removal distance and due time. The next run excludes the unchanged, already-passing level-gap case. It has not started.
• The corrected three-class fault run has the serialized private runtime slot at frozen Core c8349f9e and Package 6e3ea9ec. No result at this source pair is claimed yet.
• At Core c8349f9e and Package 6e3ea9ec, native and debug Wasm builds passed and all three classes completed thirteen quests. All three callers still failed an inventory identity assertion that compared the Available giver with the later Held loot source. The records also show the declared missing target absent before its sixty-second respawn and ordinary combat resuming. Reviewed Core 3599a0de and Package b6f0a4c compare identity within each actual held stage and bind the missing-target observation to its measured distance and due time. The three-case rerun is queued; these failures remain saved.
• Before the first persistent load run, the reviewed capture windows were set to 60 seconds for 10 bots, 90 seconds for 25 and 300 seconds for 100. The longer windows allow the bounded scheduler to move every bot through the wall route. No result at these final windows is claimed.
• The corrected three-class fault run now owns the serialized runtime at frozen Core 3599a0de and Package b6f0a4c. It excludes the unchanged, already-passing level-gap case. The exact 30-journey CI generator has also been prepared, with ten cases per class and explicit test-list checks. Generator corrections will ensure all regression jobs use the final Core and evidence-write failures cannot pass. No result from the new runtime or CI matrix is claimed yet.
• At frozen Core 3599a0de and Package b6f0a4c, Warrior and Priest each completed all 13 quests and failed the same final timing assertion. Their snapshots observed the 30-second deferral about 502 milliseconds after it was set, just outside the test allowance. Saved alternative actions occur before expiry and resumption occurs after expiry. Mage is still running. The failure remains recorded while the owner checks the assertion against the actual decision timestamp.
• The 30-journey CI generator passes source review at SHA-256 9b99ad5c71dff6d6470f139e1b8e8b127d7839bf3ec1b82d7eb940bca08b910f. It requires ten exact compiled tests per class, pins all current regression jobs to the same final Core, preserves both test and evidence-write failures, and retains the actual Wasm. YAML and shell checks pass. Actual final pins, generated workflow review and CI execution remain required.
• All three runs at Core 3599a0de and Package b6f0a4c completed their gameplay and failed the same observation-derived timing assertion. The orchestrator verified twelve runtime records, 13 rewarded quests per class and identical pre/post Wasm, SHA-256 b5e576cb8d2fe5e14e567dfb64885808dbd60523bd67ba7f7aa05e29e9aea627. Source review clears Core 49f8ff0f and Package af4d08bc: the fixture reads the actual Recovery transaction time, binds its deadline to the recorded destination and requires exactly 30 seconds. A representative replay and the final 30-case matrix remain required.
• The representative Priest at Core a2cc53ef and Package 44578182 completed all thirteen quests and recorded one successful simple GameObject use, bound to the exact Character, Core Quest row, retained objective and target. The run then failed an assertion that looked for prerequisite 783 inside the refusal text; the number is a separate field. Core acaae014 requires the actual text and keeps the numeric check. The orchestrator verified all four runtime records, identical pre/post Wasm with SHA-256 06b34ac6c76b9a148e2b62213ec87b6fda67946ccdf911e829b0e3ff8b19ac38, and the remaining completion assertions. The failed run remains recorded. A corrected runtime and final thirty-case matrix remain required.
• The orchestrator has assembled a separate final source candidate from the transfer implementation, acceptance journeys, persistent-load fixtures, imported-world fixtures and released observer pin. Conflict resolutions preserve transfer behavior and append new enum variants after existing stored variants. Independent review is in progress; no build or runtime result is claimed for this candidate.
• Independent review clears the combined acceptance Package source. The final Core now includes the actual merged command fixes and reviewed transfer changes; only the acceptance caller and released observer pin differ from the transfer base. Both final branches were rebased without changing their file contents. The prepared focused run will compile the full four-Package union once and execute Warrior, Priest, Mage and ordinary-kill level-gap cases. Its execution script is under review; no combined runtime result is claimed.
• The focused four-case execution scripts now pass source review. They require exact clean source identities, compile the full Package union once, check all four expected test names, preserve command and evidence-write failures, and compare pre/post runtime Wasm bytes. The final combined execution remains pending.
• Independent Standards review clears Core 5cd1278fee66e4a50eaf5081ed45228c72c28456 and Package ad3ade1f830d1a4b9a4578561c3c8f5cbd0dfb22. The review checks bounded ordinary-gameplay Grind, deferred Quest admissions, level-one imported staging, Frozen setup, Transfer behavior and stored field/variant order. Both actual merged command dependencies are ancestors. The prior Spec review and reviewed incremental changes remain recorded. No combined runtime result is claimed.
• The orchestrator incorporated the reviewed Transfer fixture corrections into clean Core cd08a4d9449898427a8cd8f92d42f773b49702f4 and Package 331e1413e34a2957cad4170dc6454267b93336b4. Seven source checks confirm exact Transfer caller/fixture contents and unchanged acceptance changes in both repositories. The corrected acceptance caller is unchanged. This is source composition only; final combined runtime remains pending.
• The orchestrator composed the latest partition-schema and debug fixture corrections into Core 8ca03639 and Package f19fb96. Both worktrees are clean. Acceptance-only patches remain byte-for-byte identical to the previous reviewed deltas, ready to apply to the actual PB-010 merges while retaining their CI changes. This is source composition only.
• Read-only review confirms both acceptance-only patches apply cleanly to Core 6fc104e5 and Package fc45fbea. They preserve Transfer fixtures, party history, Assist, debug registrations and Runner field order. The new stored Grind variants append after existing tags. Integration will use fresh branches from the actual Transfer merges and these patches only; it will retain the new Package CI workflow. This is source preparation, with final runtime still pending.
• Headless Client PR17 merged normally at 68f4a0dec63feaba330efbb15f7730ccc4c1d805 and is published as v0.1.0-alpha.13. Local formatting, Clippy, 87 tests and release builds pass; Linux, macOS and coverage CI pass. The scenario sends actual party addon commands, timed human movement and AreaTriggers, retaining packet and reply evidence. This establishes the client driver; the companion acceptance callers have not run.
• Corrected action lifecycle source is frozen at Core 491a6600 and Package bf5a63ad. The manifest binds 75 cells to 44 distinct source cases and two explicit Resurrect dispositions. Independent standards and specification review is clear; compilation and runtime remain pending. The execution capture runner passes thirteen focused checks, including cleanup, time limits and accurate failure records.
• The final companion source at Core29c5acb3 and Package9428cd25 passes independent standards and specification review. Stay uses its committed anchor and normal approach distance, then measures later drift from the stopped position. The wound preserves location; healing receipts bind to the fault time; inventory checks retain exact identity and counts while accounting for ordinary durability loss. No companion runtime pass is claimed.
• The autonomous, lifecycle and companion deltas are composed at provisional Core0f00529b and Packageccc6ed40. The source checker verifies 15 actions, 75 cells and 44 named cases, with execution checking explicitly disabled. A focused run is being prepared for four autonomous cases, fourteen new lifecycle cases and two companion cases. Actual PB010 merges, runtime evidence and canonical CI remain required.
• The reviewed autonomous, lifecycle and companion changes are composed at Core f20d0888b76a5fddc857ac717f31258bbe8b4445 and Package 52a72d5ccffc2721e22ecfcc0d47355fafc6c4a2. Core now uses the actual PB010 merge; Package inherits the final PB010 CI correction. These sources are clean and frozen for twenty local cases: four autonomous journeys, fourteen lifecycle cases and two companion scenarios. No result is claimed before execution.
• Draft Core PR505 is open for CI at the reviewed composition. The first twenty-case launcher stopped while recording Package ownership before compilation or private runtime began. Its manifest treated uninstalled Package links as Core files. The correction records declared ownership before setup and checks the actual links after setup. The failed evidence directory is retained; no scenario pass is claimed.
• Core PR505 CI found JSON parser type and query-lifetime errors in the new companion caller, plus two autonomous test lint findings. Core52844b17 corrects them without changing gameplay. Two local compilation attempts then identified both private threat-observer reads that require a narrow Package API exemption. Package3f227f32 places the exact function import behind that documented fixture exemption. Each failed attempt retains all twenty cases as not run, with source and cleanup checks passing. The corrected full-union attempt is next.
• The fifth composed attempt at Core52844b17 and Package5dbb6b29 passes both native Module compilations and the fresh full-union Wasm build. It has begun the first autonomous Warrior journey through ordinary scheduled decisions. Four earlier setup or compilation attempts remain recorded as no runtime. No scenario pass is claimed until its exact result is recorded.
• At Core52844b17 and Package5dbb6b29, Warrior783001000, Priest783005005, Mage783008009 and the ordinary-kill level-gap journey each pass their exact case. The lifecycle group exposes failures before its expected movement or root assertions, and a fixture spell collision. All cases continue so the complete failed-run evidence is retained. Source diagnosis and a corrected rerun are required; the twenty-case suite is not a pass.
• Core acceptance PR505 passes formatting, lint, unit, build, Wasm and durable CI at52844b17. It is ready for review and remains unmerged while relevant acceptance results are pending. The persistent 10, 25 and 100 bot execution plan passed source review, with a separately planned build and the pinned gzip observer. No final load result is claimed.
• The focused composed run finished with twenty exact cases executed: five pass and fifteen fail. All four autonomous cases and the resurrection cancellation case pass. Thirteen lifecycle cases fail and both companion cases stop during obstacle setup with an empty obstruction-grid bounds panic. Result capture, Wasm identity, cleanup and clean-source checks pass. Failed results remain preserved. Corrections are in isolated worktrees and no complete lifecycle, companion or release pass is claimed.
• The corrected lifecycle caller passes independent and orchestrator source review. It completes ordinary provisioning before ReturnHome, binds Quest purpose through retained identity and Recovery, records blocked movement without inventing a spline, and avoids conflicting fixture ownership. Companion obstacle setup now accepts an empty obstruction grid and uses its declared base height. The lifecycle checker binds exact Rust test targets and passes seven focused checks. The composed Core3b387fa7 and Package71baac45 pair is ready for fourteen lifecycle and two companion cases; no corrected runtime pass is claimed yet.
• The corrected focused run at Core3b387fa7 and Package71baac45 executes all sixteen declared cases. RecoveryPosition expiry, resurrection cancellation and admitted Transfer with an unrelated expired objective pass. Eleven lifecycle cases and both companion cases fail. The fixed companion route waits for an absent wire command result; the restart scenario hits the CLI permanent-aura timestamp formatter while saving evidence. Native builds, exact case listings, fresh Wasm, complete capture, source checks and cleanup pass. Wasm SHA2568b7fd6ae6fe0180a8b89ac9d57066f7604929af7d7159d9b1f4f3c8b6125299c is identical before and after the run. This is a failed acceptance run.
• A one-case diagnostic at Core14bf3459 and Package71baac45 saves the Account, live body and Character after authenticated Headless Client login. Account and body have the current bound identity; Character retains zero. The diagnostic fails, and its first identity assertion also needs the nested SATS identity unwrapped. Core8edee202 corrects login ownership, checks all three identities before companion commands, and saves permanent-aura evidence without the unrenderable expiry timestamp. Two independent source reviews pass; corrected runtime remains pending.
• The corrected composition at Core dd385792953a48d41b128b1b55383516ba8511f9 and Package 1c2833353f37a4bdabc97ef48b255630f911d049 uses the actual merged PB-010 predecessors. Source reviews and seven checker tests pass. The matrix resolves 15 actions, 75 cells, 44 source cases and two Resurrection exclusions. ReturnHome owns its Transfer and TransferPosition deadline before admission. The following record contains the completed runtime result.
• The focused run at Core dd385792 and Package 1c283335 executed all sixteen cases exactly once. Six lifecycle cases passed; eight lifecycle cases and both companion cases failed. Capture, case accounting, unchanged source checks and cleanup passed. The fresh four-Package Wasm was 9,553,466 bytes with SHA-256 8c93192d95c8a98b22570aa6ce911f5a4ed42ddade03fe45eab7d77e1e3c60fd before and after execution. The companion cases passed their earlier login stage, then stopped at unsupported SQL identifier quoting and a parser that split the quoted 1|Applied reply. Neither companion scenario is accepted yet.
• Core fc41d790 corrects both SQL reads and preserves separators inside quoted result cells. Independent source review passes. All six shared parser and polling tests pass with exact source and executable checks. Two earlier evidence-parser attempts are retained; their Cargo tests passed, but the wrapper did not recognize the should-panic result decoration. Core bae6853b also makes each companion snapshot lookup name its GUID field. Runtime with these corrections is pending.
• Both companion retries at Core bae6853b and Package 1c283335 failed. The fixed route completed Follow and committed Stay, then its SATS parser matched the y in stay instead of the coordinate field. The command-restart case recovered the applied intent, then its new Headless Client hit ACCOUNT_IN_USE because the crashed Gateway still held an unexpired 60-second Account Claim. Both cases executed once; capture, source identity and cleanup passed. Wasm SHA-256 e67ca76527a043f468a098dae7e38d69a9f064e044199abbd9a8080eece2db5d matched before and after execution. Core 46c9fe6f corrects complete field-name matching and has independent source review. Reconnect must observe normal claim expiry without changing the ownership Gate.
• The next reviewed composition is frozen at Core 44495802b2e84d15c9f17fdf6f4b5e0226446159 and Package d552add5c575ba6537e2e1b583800c3ea494a408. It includes the lifecycle setup corrections, exact SATS field parsing and a reconnect caller that waits for natural Account Claim expiry, then checks the new generation. Both review axes pass. The source matrix resolves 15 actions, 75 cells, 44 cases and two Resurrection exclusions. The private run is starting fourteen lifecycle cases, two companion scenarios and two separate ordinary parser checks. Core PR #505 is updated and CI is running. No new runtime pass is claimed.
• The completed run at Core 44495802 and Package d552add5 passed twelve of fourteen lifecycle cases and both ordinary SATS parser checks. Both companion cases failed: the fixed-route pull did not finish, and the restart comparison observed a changed Companion objective deadline. The Quest failures concern a retained-source read before planning and a casting-position setup whose first leg already reached casting range. All eighteen checks executed once. Capture, result accounting, source identity and owned-process cleanup passed. Wasm SHA-256 68abd29040f728042cc8eae3030398e990e870c7ca41d1f35d0369978e110a90 matched before and after execution. Natural Account Claim expiry and replacement generation passed through real login. The failed results remain retained; no companion acceptance is claimed.
• The next composition is frozen at Core 7b74db85185f9c149d0d0be3fed2639aa21ae793 and Collection 2a290fb7e5c4d0b2eb95c77ba557b4387ed72c01. Independent Standards and Spec reviews pass. Recovery now preserves the unlimited Companion objective deadline when real progress occurs. The two Quest fixtures use the actual loot source and keep a casting-position move pending after observed progress. The companion fixture aligns the three declared spawn homes with their route positions. The previous pull timeout saved no terminal combat state, so its cause remains unproved; new progress and timeout snapshots retain that evidence. The focused run executes the two failed Quest cases, finite ReturnHome expiry, and both companion scenarios. All fourteen lifecycle test names still pass compiled-list accounting. The source matrix still resolves fifteen actions, seventy-five cells and forty-four cases. Core PR #505 is updated; its prior revision passed every required CI check. No result is claimed for the new runtime or CI.
• The completed focused run at Core 7b74db85 and Collection 2a290fb7 passed all three lifecycle cases and failed both companion scenarios. Every case ran once; capture, result accounting, source identity, Wasm identity and owned cleanup passed. Wasm SHA-256 ff309170ffb19477fec7bc4b22b7ab94be8c88ae46194bd974a2b409a2608a00 matched before and after runtime. The fixed-route target fell from 1000 to 420 health through one Mage while the Warrior and second Mage remained at clipped routes beside the wall. Both stalled bots had zero accumulated stall time and no retained route observation. Recovery removed their attempts whenever the shared target lost health, before checking their own movement. The proposed correction preserves approach accounting while the previous action is movement, while target death still ends the fight. The restart case passed the retained Companion deadline comparison, then its new Gateway refused client login because Character ownership did not match the Account. This is under diagnosis. No companion acceptance is claimed.
• Independent review cleared the stuck-approach correction at Collection 96885347. It is composed with an import cleanup at 23e1a849; no runtime pass is claimed yet. The reconnect diagnosis identifies a production ownership mismatch: Transfer creates a destination shadow Account, but both Gateway admission and Module fencing require its local placeholder name to equal the authenticated Account name. A correction is in progress with independent security review. The cast-evidence diagnosis is separate: Fireball damage occurs at its later projectile impact, while the private observer copies the earlier zero-damage cast-GO event. That observer is being corrected to retain authoritative impact identity and damage. Core PR #505 passed all required CI at 7b74db85, including the durable suite; this does not replace the two failed companion cases.
• The Recovery and projectile-evidence corrections are composed at Core 108e02d8 and Collection 38317707 with clean working trees. Independent review cleared the exact source. Projectile receipts now retain authoritative impact identity and damage only for one unambiguous cast match. Every pull boundary checks failure codes, unique impact identity and event order. The bounded receipt budget is 256, covering three declared 1000-health targets at 20 damage per Fireball plus support casts. The companion topology now records and requires different route Account ids on Realm-core and the source World Shard, so real login and restart exercise local-id to authenticated-name to Realm-id resolution. No new runtime pass is claimed. The separately reviewed next execution script declares exactly seven Module ownership cases, one ordinary Gateway ownership case and both companion cases; process and evidence cleanup failures remain failures.
• The Account and companion retry at Core 06b57d33 and Collection 38317707 stopped during the fresh debug Wasm build. CompanionCastReceipt lacked Clone. All eleven declared cases are recorded as not run; owned build and process cleanup completed. The reviewed correction adds Clone and uses a mutable slice for retained receipts. No gameplay pass is claimed.
• Core PR #507 isolates retained Account ownership after Transfer. Its initial CI passed build, formatting, ordinary tests and Wasm. Six of eight durable ownership cases passed. The other two exposed test defects: successful logout legitimately persists live Character fields, and a single fixture INSERT exceeded the SQL request limit. A needless borrow also failed Clippy. Corrections and repeat validation are pending.
• The corrected eleven-case private retry started at Core 46784682 and Collection 272f0a69. All seventeen composed files match reviewed source. The action matrix still resolves fifteen actions, seventy-five cells and forty-four source cases. The Account test binary compiled; fresh debug Wasm and runtime results remain pending. Core PR #507 is repeating CI at 48e8d0a3.
• At Core 46784682 and Collection 272f0a69, all eight exact Module Account ownership cases passed, including Account switching, stale-body cleanup, bounded reaping and 4,097-row overflow refusal. Fresh pre/post Wasm bytes matched, with SHA-256 f08db378f49bf3b976a6a827bf9c790fed551215bb6543d661b4860bf9c258a1. Evidence capture, source checks and owned cleanup passed. The run stopped when its Gateway command selected --lib for a binary-only crate; one Gateway case and both companion cases are explicitly recorded as not run.
• The independently reviewed replacement command selects the declared Gateway binary. Its manifest contains only the three unrun cases, preserving the eight passing results at the same unchanged sources. Core PR #507 now has the reviewed documentation correction at 15d8858f; all nine non-Markdown PR files match the private runtime source byte for byte. Required checks are still running.
• The three-case retry at Core 46784682 and Collection 272f0a69 passed the Gateway Account ownership case. Both companion scenarios failed later: the fixed route completed its first pull before an invalid attempt to associate cast completion and projectile impacts; restart passed the prior login failure before an exact JSON comparison found one f64 rounding step in serialized f32 movement. Evidence capture, source and Wasm checks, and owned cleanup all passed. These remain failed companion results.
• Core #507 merged normally as 40f91599b92ac06d8230bb5efdda4cd598c5a20d at 15:04:33 UTC after all required checks and review passed. Its nine non-Markdown source files match the eight passing Module ownership cases and passing Gateway case. Core #505 is rebased onto that actual merge and now has head e1136dd3930e1c510252f13969213cb3bf45cf60.
• The reviewed acceptance correction retains cast completion and projectile impacts separately because their public events have no one-to-one identity. Each retained impact preserves the actual source event id, declared participants and targets, damage and timestamps. Command evidence now compares all fields against the exact serialized request while preserving requested f32 coordinates. Independent Standards and Spec review passes for Core 05669908 and Collection 235b4413, composed at Core e1136dd3 and Collection 8d6bcb07.
• The current three-case run at Core e1136dd3 and Collection 8d6bcb07 passed serialized_move_command_preserves_requested_f32_coordinates. The fixed route advanced beyond the prior impact evidence defect, then failed after 178.65 seconds because scheduled fault 0 never became due. The restart scenario is running. Final capture and cleanup results are pending, and no companion acceptance pass is claimed.
• The completed run at Core e1136dd3 and Collection 8d6bcb07 passed command serialization and failed both companion cases. Fixed route failed after 178.65 seconds; restart reached its final inventory assertion and failed after 208.73 seconds. Capture, case accounting, source checks, Wasm equality and owned cleanup passed. Pre/post Wasm SHA-256 is 85ac0efc2f54d838d98f5c179cc37e4e10ac7ea204af0ab73d5aa766ce0bb18f.
• Fault 0 was already due by 56.225588 seconds. All 291 requests failed because Fortitude was absent and the fixture tried to cancel it. Atomic rollback preserved applied_micros at zero. The corrected Package conditionally invokes normal cancellation and requires the buff absent. The Core caller saves attempts before failure and retries only the exact declared not-due Refusal. The later real heal and Fortitude cast assertions remain.
• The restart replay preserves all 26 item identities and stack counts. Its provisioning cursors and histories advanced through ordinary maintenance. Four human-owned items received their normal authenticated identity on first login, after the old baseline. The caller also omitted the supply catalog, so all four companions lacked the bandages required by the next assertion. Reviewed Core 80a1905c stages the existing catalog, takes the baseline after login, requires exact supplies before the route, compares stable profile facts and retains exact item checks. The equipment roster now includes the human leader whose items were already measured. Repeat runtime evidence is pending.
• The next private run is prepared at reviewed Core 80a1905c3faee890b5bb38d7c33e405fb5c5e3a0 and Collection 4c61b2bb4c217d1a782d65a2e1acb0577b6f7886. All eighteen retained source files match their reviewed revisions. The source matrix resolves fifteen actions, seventy-five cells and forty-four cases. The independently reviewed two-case wrapper retains fresh Wasm, exact Package identities, captures and owned cleanup. Both companion outcomes are pending. Core #505 now names this pair.
• At Core 80a1905c and Collection 4c61b2bb, fresh Wasm and the companion target compiled and both declared cases executed. Fixed route failed after 147.91 seconds while waiting for the Priest healing predicate. Restart failed after 209.13 seconds at its rewarded Quest comparison, after all inventory assertions passed. Capture, case accounting, source identity, Wasm equality and owned cleanup passed. Pre/post Wasm SHA-256 is 9c5810e6503a15c2e8d6586b741dcdb017d6b0d91d461ba0564b2762d0876708.
• The restart snapshots contain exactly one row each for Warrior Quests 50910 and 50911. Their only changed field is the local row id, 2 to 4097 and 3 to 4098. Core declares these ids reminted on Transfer. The correction compares every retained field except that id and requires one exact Character/Quest row. It preserves the full row comparison across a Module restart without Transfer. The next correction also captures full state when a companion wait expires, with the same predicate, deadline and cadence. Healing diagnosis and repeat runtime remain pending.
• The saved pre-fault state has Priest power 0/0 and eight Fortitude training refusals. It cannot satisfy the required real casts; no post-fault snapshot exists to resolve every timeout condition. Reviewed Core 25aa28d1 calls the existing private trainer-offering setup before ordinary provisioning and declares the existing 100-point fixture mana pool before begin. It does not create healing, auras, casts, movement or receipts. The next two-case run uses the unchanged reviewed wrapper at Core 25aa28d1 with Collection 4c61b2bb. All eighteen source identities and the action matrix pass source checks. Actual companion results remain pending.
• At Core 25aa28d1 and Collection 4c61b2bb, the restart/Transfer companion case passed in 192.66 seconds. The fixed route failed after 152.67 seconds. Its saved timeout state has both required Fortitude auras and 9.64246 yards of Priest movement, but no required 2050 heal receipt; Priest power and maximum power are zero again. Diagnosis is tracing the owning stat inputs. Both cases executed once. Capture, case accounting, source identity, Wasm equality and owned cleanup passed. The Wasm SHA-256 remains 9c5810e6503a15c2e8d6586b741dcdb017d6b0d91d461ba0564b2762d0876708; timeout snapshot SHA-256 is b5fa091aaa4eac1be2148ff10e810af595b8ee7ebbc7943cc44b0f87396ee57f. The passing restart case does not complete the fixed-route requirement.
• The timeout diagnosis found the exact mana reset: Priest power was 100/100 before self-Fortitude and 0/0 afterward. The normal aura stat owner derives both health and mana from class-stat rows, which this private fixture had omitted. Reviewed Core a4f0f3ca seeds only the declared level-5 Priest inputs before spawn in both World Shards, preserving 120 health and zero base attributes while deriving 100 mana. It removes the direct live-body mana assignment and records the stat rows in snapshots. Nineteen composed source files match their reviewed revisions and the source matrix passes. Both companion cases are running again; their new outcomes remain pending.
• At Core a4f0f3ca and Collection 4c61b2bb, restart/Transfer passed in 208.61 seconds and the fixed route failed after 151.66 seconds. All capture, source, Wasm and cleanup checks passed. The Priest retained 100 maximum mana and healed the Mage for 50 through a real 2050 cast. The human leader recovered above that spell's 50 percent threshold while the Priest repaired line of sight, so the second required heal was absent. Reviewed Collection a0bb92e changes the declared wounds from 35 to 10 percent health, retaining normal regeneration and every heal, buff and movement assertion. Core db4951dc extracts the unchanged stat setup into a named helper and passes Clippy. The next run executes only the changed fixed route. Its wrapper retains exact case discovery and accounting, fresh Wasm, source identities, captures and owned cleanup. Nineteen source files match their reviewed revisions and the source matrix passes; new runtime results remain pending.
• At Core db4951dc and Collection a0bb92e, the fixed route passed both required real Lesser Heals. The Priest healed the leader for 50 to 62 health and the Mage for 50 to 86 health, with distinct scheduled casts and source-event receipts. Fortitude and physical movement also passed. The case then failed after 189.79 seconds because the Mage did not assist the named Priest. The saved state has a valid leader-issued Target order on the Priest and the correct Runner fight target, but its live target field is zero. Assist reads only that live field and records TargetUnavailable. The proposed correction recognizes a current authorized Target order on a same-party Cohort companion and retains the ordinary target Gates. A new durable regression covers current and replaced targets, crowd control, death, partition changes, and ended controller authority. The complete route remains pending.
• Reviewed Core 056a4132 and Collection 9d481bf4 correct Assist authority. A same-party, sessionless Cohort companion can supply its current leader-issued Target order before attacking. Core still checks the exact target, and a refused target never selects another fight. The durable regression covers replacement, crowd control, death, partition changes, controller suppression and an Account claim on the named companion. It proves the retained order loses pull authority after that claim while the assisting Mage remains eligible. Both review axes pass; report SHA-256 ecaaed02175543af4846f243a90fc25c7913fbd66c4ffe8e751ac2c02b55051b. Twenty-three composed files match reviewed revisions, and the source matrix remains fifteen actions, seventy-five cells and forty-four cases. The next private run executes three exact order cases and the full fixed route with a fresh four-Package Wasm. Runtime results remain pending.
• At Core 056a4132 and Collection 9d481bf4, all three focused Module order cases passed in 12.11, 14.01 and 10.26 seconds. The full route passed healing, Fortitude, movement, Assist and crowd-control cancellation, then failed after 187.73 seconds at the second-pull top-threat assertion. All capture, source, Wasm and cleanup checks passed. Wasm SHA-256 is 6edcff9d816ba9e320273d61978d09edb11dd231724b5da852901ca00b3bb167. The Warrior remained level 5, made no casts, and its provisioning history correctly refused Taunt until level 10. All 408 live second-pull samples targeted the Mage. This is a sustained toolkit limitation, not a transient receipt miss.
• Reviewed Core 02405a81 declares the synthetic companion tank at level 10 before ordinary provisioning and before the route begins. The saved setup must prove both Character and body level 10, the unchanged Taunt level requirement, and exactly one normally learned Taunt. The source retains the authoritative top-threat assertion and adds spellbook, spell-header and raw-threat capture. Other party levels and imported progression are unchanged. Both review axes pass; report SHA-256 7c9c56ef11685c58ef7a14366e91ff0e7bc21f11e7e4b77ea158349e4b8555bd. Twenty-three source files match their reviewed revisions and the source matrix passes. Both shared-setup companion cases are running with unchanged Collection 9d481bf4; results remain pending.
• At Core 02405a81 and Collection 9d481bf4, both scenarios stopped at the preflight Taunt check in 10.04 and 9.96 seconds. The saved Warrior Character and body were level 10, but training correctly refused Taunt because no Warrior trainer offered it. The Fortitude offering makes this catalog nonempty, so the empty-catalog demo admission used by the separate role test cannot apply. Capture, source, Wasm and cleanup checks all passed. No route gameplay began in either case.
• Reviewed Core f384cfb1 and Collection 85176fda stage one private Warrior trainer and one direct Taunt offering at its real level-10 requirement. The existing LevelUp provisioning hook arms the current level; ordinary reconciliation learns the spell. No spellbook row, threat value, spell rule or combat result is assigned by this setup. Before begin, the saved state must prove Character/body level 10, unchanged Taunt requirement, provisioning level 10 and exactly one learned Taunt. The snapshots retain trainer inputs and raw threat. Both review axes pass; report SHA-256 9fe3950f484d2206393f52c0d31070f1cb1ecd221405204f1250e368eeb94053. Twenty-four composed files match reviewed revisions and the source matrix passes. Both companion scenarios are running again; actual results remain pending.
• At Core f384cfb1 and Collection 85176fda, restart, Transfer and lost-acknowledgement recovery passed in 210.48 seconds. The fixed route passed its Taunt prerequisites, second-pull tank threat and death/regrouping checks, then failed after 263.47 seconds because the third enemy remained alive at 44 health after the 60-second combat wait. Captured observations show continuing damage, so the cause remains under diagnosis. Both cases were accounted for and capture, source, Wasm and cleanup checks passed. Result report SHA-256 28adc2f3cae4b6723cdfaf30a92161bc514e1053478a5658da01759c602ac62f. This is one passing scenario and one failing scenario; PB-011 remains incomplete.
• The third-pull diagnosis found the caller issued Target while the resurrected Mage was still 235.70 yards from the human leader. Its first damage arrived about 31 seconds into the fight; both Mages continued acting and the enemy continued losing health. Core 78290e0b now waits for ordinary Follow to reach the existing 3.05-yard arrival radius and saves that regroup boundary before issuing the next Target. The original physical-progress, recovery-identity and combat assertions remain. The enemy health and 60-second combat budget are unchanged. Independent Standards and Spec review passes. Twenty-four source files match their reviewed revisions and the source matrix passes. A fresh private retry runs only the changed fixed-route case; the unchanged restart case retains its preceding pass.
• At Core 78290e0b and Collection 85176fda, the added regroup check passed with the resurrected Mage 3.011 yards from the leader. The fixed route still failed after 298.30 seconds when its third fight exceeded the unchanged 60-second wait. The final snapshot has the enemy alive at 149 of 1000 health, forty real Mage impacts and authoritative Warrior top threat. All capture, case accounting, source, Wasm and cleanup checks passed. Result report SHA-256 95716c05079b3cc97ab6e37a4520bc5fd41708d2423a1f82e8afdb3fb4c8bd6a. Earlier snapshots also show repeated movement cancellation and a NoMovement result before that fight. Those decisions and the combat budget are under investigation; the route has not passed.
• Source review identified a controller defect before the third fight: both Assist and Target Refusal branches called full cancellation before Recovery could observe owned physical movement. An unchanged dead-target order could repeatedly cancel Follow or friendly casting, leading to false NoMovement despite earlier displacement. Collection eda8a95 removes those two calls. Existing later cleanup still cancels obsolete offensive melee, casts and approaches using current selected targets. Both review axes pass; review SHA-256 effe5b728218c78bfe8740e287bb8da9e56101fbc2078d5eaef4eddeec8e287f. Core ca6cb3cf adds two reviewed durable regressions for following through repeated Refusals and completing an exact Lesser Heal cast under a refused Target. The first run deliberately uses the old Collection 85176fda. The corrected code has not yet passed runtime validation.
• The first regression attempt stopped at native compilation because two numeric comparisons needed explicit u64 parsing. Both behavior cases were not run. Wasm, source, capture and cleanup checks passed; no behavioral failure or success is claimed. Core ce5603e3 adds the reviewed type annotations and a fresh retry now runs the same two cases against old Collection 85176fda. The separately reviewed controller correction eda8a95 remains outside that run. A later affected-suite run will also use a bounded 90-second pull wait, accounting for the declared 1000-health enemy, two 20-damage Mages and measured physical approach. Actual combat assertions and enemy health remain unchanged.
• At Core ce5603e3 and old Collection 85176fda, both new regressions compiled and ran. They failed their final dead-target assertions because the reserved creature decayed and respawned. The Follow snapshot also recorded NoMovement and no retained movement progress, consistent with the controller defect. The friendly cast completed after the target became live, so it did not prove behavior under repeated Refusals. These failures are not accepted as the intended regression comparison. All capture, source, Wasm and cleanup checks passed. Result report SHA-256 8cb4cf937b090f7fcdcedecc943460302c61ca4761b05b23e1d98c1f754c2699. The correction belongs in the private fixture: disarm the reserved dead creature lifecycle timers once, retaining the real controller, movement and casting behavior.
• The stable dead-target correction passes both review axes. Collection 2fb8cb95 contains only that fixture change over 85176fda; its controller is byte-identical to the old code. The private fixture parks both corpse decay and respawn timers for the same reserved target, without changing normal creature lifecycle or bot behavior. Review SHA-256 8c94e72d5ba81219f07d5da9048bf79903e3ae1378812d3aff8f8d4b5f7e4146. Twenty-four composed files match reviewed revisions and the source matrix passes. The two regressions are running at Core ce5603e3 against this old-controller comparison. The controller correction remains outside the run.
• The stable-fixture comparison reproduced both intended defects at Core ce5603e3 and old-controller Collection 2fb8cb95. Follow failed after 71.99 seconds with no retained movement progress, a NoMovement result and the same active refused Assist order. Lesser Heal failed after 13.96 seconds: the first scheduled cast never completed, the current scheduled cast was already 59, and history showed repeated cancellation. Both targets retained TargetDead. All source, Wasm, capture and cleanup checks passed. Result report SHA-256 04cdca1cad090524a79f89ceafdbb9ba62341fec4edcb6eb33f3dac3d10c0f50. Collection f33e876a now adds only the reviewed six-line controller deletion to that comparison. Its complete source tree equals reviewed 9b66d3ba. Twenty-four source files match their reviewed revisions. The next run executes both regressions, three existing order-safety cases and both complete companion scenarios; no corrected-runtime pass is claimed yet.
• At Core ce5603e3 and corrected Collection f33e876a, both regressions passed: moving-member Follow in 63.28 seconds and completion of the exact Lesser Heal in 7.56 seconds. The three existing order-safety cases passed in 12.06, 14.17 and 12.60 seconds. These cover the named companion current Target order, ordinary Assist authority, and exact target admission with cancellation of obsolete controlled work. The same run is compiling and executing both full Gateway companion scenarios next. Final capture and cleanup checks remain pending until that run finishes.
• The corrected run passed all five Module order cases and restart/Transfer in 209.36 seconds. The fixed route failed after 193.08 seconds because it retained no authoritative second-enemy top-threat receipt. The enemy died, but this does not prove the tank criterion. Source review found that threat observations only updated an existing damage receipt, so Taunt without melee damage could leave no receipt. A separate real behavior defect is visible across four saved phases: the Warrior stayed 2.654 yards from the enemy but faced 148.27 degrees away while accepting melee attacks. Core correctly refuses a Character swing at a target behind it. An earlier passing run happened to turn through movement. The bot needs explicit facing for its admitted melee target. Both corrections are under review; neither the route nor the rewrite is complete.
• The exact-target facing regression reproduced the intended failure at Core 59005dd4 and Collection aaa0787c. The remote request returned Armed and retained its movement leg. At melee range, AlreadyArmed left orientation at zero while the target was behind the Warrior; no positive hit occurred and the swing clock stayed zero. The case failed at the facing assertion after 66.40 seconds. Capture, exact case accounting, source, Wasm and cleanup checks passed. Result report SHA-256 9b891fa9785e696b9c1e32d935a6286b925551bbd42c01b936496718d238a6d5. The correction uses the existing Core facing rule, preserves active movement and swing timing, and sends the existing facing update. Source review passes at Core 28ab815f. The same test and Package now run against that correction; no corrected-runtime pass is claimed yet.
• The same facing regression passed at Core 28ab815f and Collection aaa0787c. The remote accepted attack retained its active movement exactly. At melee range, AlreadyArmed turned the Warrior from zero to pi and produced a new exact-target hit for 6 damage. After the target moved across it, another AlreadyArmed turned it back to zero, retained the immediate swing clock exactly, and produced another new hit for 3 damage. All final case, capture, source, Wasm, process and cleanup checks passed. Result report SHA-256 db4cc3aa43995bf937511f2c5d3f200b283c75a8427adb3c36eef98f58b10c63. The five existing order cases and both complete companion scenarios now run at this same pair. This focused pass does not complete the full route or release acceptance.
• At Core 28ab815f and Collection aaa0787c, the complete human-plus-four companion route passed in 372.31 seconds. It progressed through the previously failing second-fight tank assertion, death and resurrection, physical regrouping, the third fight, recovery and remaining instance assertions. All five affected Module order cases also passed. The restart/Transfer/lost-acknowledgement case is running next. This is six completed passes out of seven; final batch capture, source and cleanup checks remain pending.
• The final batch passed all seven cases at Core 28ab815f and Collection aaa0787c. Restart/Transfer/lost-acknowledgement recovery passed in 209.62 seconds, completing the earlier six passes. Every capture, exact case, source, Wasm, process and cleanup check passed. Independent review of the raw route and restart evidence is clear, report SHA-256 e574909cb922385f76ef78fc583fdcc584ab7ab3047fd7c521e70f444917e531. All Core PR checks now pass, including the durable job, with no unresolved review threads. Persistent-load measurements remain required before the Core merge.
• Two persistent-load invocations stopped before compiling Wasm or measuring: cleanup and build commands depended on the caller directory. No private load Shard started. The reviewed correction explicitly runs both commands in the Core checkout. Mocks confirmed cleanup failure handling, the exact build arguments, visibility of all four Packages and restoration of the source tree after success and failure. A complete directory-dependency review found no other affected command. The retry uses new evidence directories and the same tested Core and Package. Its wrapper SHA-256 is daeeaa0cb022744e25fbe64accedc1d2b5895529623e1d3594c9375df34d4acf. The fixed load limits and post-build disk-space check are unchanged.
• The persistent-load batch completed at Core 28ab815f and Collection aaa0787c. Ten bots passed: all progressed, 595 decisions, 585 completed movement legs and maximum scheduler lag of 5,779 microseconds. The 25-bot path check failed on an authoritative 206-millisecond spline that crossed a blocked cell corner by about 0.088 yards. Exact intersection and the retained spline confirm the navigation sampling defect. At 100 bots, all received decisions but none moved: 100 nearby Character bodies exhausted a 96-row optional Grind search, and its refusal selected Hold above the retained ReturnHome objective. Raising the scan limit or weakening the path check is not the proposed fix. All capture, source, process and cleanup checks passed. Load acceptance remains failed. Result report SHA-256 8266a26bff38b8ce011c14576659d70ce380a06c0c25788bc0a7c6a033dab70f; navigation diagnosis SHA-256 f0dbeb5be9351c5bf61a40fe4b046f2bb6227e04792ceba688c0910a5b6143e6. Root owns navigation; the Package owner is correcting optional-search priority. Both require regressions and another measured run before merge.
• The navigation correction has independent failing reproductions for two defects: a shortcut through a blocked cell corner and removal of a final waypoint that still needed to be kept. At reviewed Core c0762cff, both regressions pass, as do exact-corner and neighboring-f32 cell-edge cases, all 176 shared tests, formatting and Clippy. Every segment check uses independently worked rectangle bounds rather than the route algorithm. Source review report SHA-256 b8c13705af3efff63109dd9172b66e29f1d5ae47711b4058b490a0ea69bd2d98. Core PR505 now contains these changes at 248c78dd and the ReturnHome runtime regression. The Package correction permits a new optional Grind search at home and preserves an existing Grind target during an excursion. Read bounds and decision priorities are unchanged. It passed independent review, report SHA-256 76c95bbf2a44818461bc3d11b3ceae5cbda8cc993c0480368d850b986efc5516. The runtime comparison is running first against the unchanged Package aaa0787c. Corrected persistent load and companion acceptance remain pending; no new load pass is claimed.
• The same ReturnHome caller failed against Core 248c78dd and old Collection aaa0787c, then passed against the same Core and corrected Collection 7d8859e8. The failing case retained ReturnHome but selected Hold(Grind) and created no movement. The corrected case retained the same purpose, selected Move(Home), owned its matching Foreground Action and emitted a 1,000-millisecond Core spline from x=1200 to x=1207. Its failure list was empty. All source, build, Wasm, capture and cleanup checks passed. Independent failing-run review SHA-256 4ae1fce28a2c86edeb85ce52f1f6381b64322e9227e251939d7ce0f6d2275cc2; corrected-run review SHA-256 da803edd243e40c668cbf18158e225fe18e29462489bba132d2117d2fa9aafe3. The complete seven-case companion and order batch is now rerunning at 248c78dd/7d8859e8. This focused result does not establish persistent-load capacity or 100-bot autonomous combat capacity. The required 25-bot imported-world hour remains pending.
• At Core 248c78dd and Collection 7d8859e8, all five order cases and the complete fixed companion route passed. The restart/Transfer composition case failed before instance entry: four party members transferred, but Mage One remained on map 0 with no Transfer Intent. Saved geometry shows its route planner sees a clear line while the movement Gate meets a synthetic obstruction. The fixture walk mask spans about 2.6 yards while the obstruction spans about 9.4 yards. Both the old sampled check and corrected exact traversal accept this terminal line, so the evidence does not justify reverting navigation. All source, Wasm, capture and cleanup checks passed. Runtime report SHA-256 9a7a31cf74b9387ab0a55dc02f3edf226aa1e412c3f659d0c4bcbef75bb399fa; geometry diagnosis SHA-256 4365579fc7e97b1e3e6d05661bd0775e509b3bbe97bdecbb00731a42095e5c50. The fixture correction must preserve collision Gates and acceptance deadlines. Load remains pending. Core PR505 passes all CI checks at 248c78dd but is not merged.
• Collection 175c3ffc corrects only the synthetic companion repair wall. Its walk mask derives from the same obstruction footprint, with two walk cells of clearance on each side. Existing line-of-sight, movement, Transfer and deadline checks remain unchanged. Independent Standards and Spec review is clear, SHA-256 5e5a546e83f26e1859bb6fb310a7aab9a119c9c5f0987a9e882fed69dc8f2c51. The complete seven-case batch is rerunning at Core 248c78dd and Collection 175c3ffc. Persistent load remains pending; no configured realm operation has occurred.
• The corrected-wall attempt at Core 248c78dd and Collection 175c3ffc passed all five Module command scenarios. Gateway linking then exhausted temporary storage, and neither companion case started. Quota also interrupted capture, leaving eight empty capture or summary files. The failure remains recorded; it cannot count as a complete acceptance run. Operational cleanup passed, both source trees are clean and no task-owned process remains. Cleanup receipt SHA-256 7d51f912991e5794ca44cbf237b6061d96760cfae48bfe8f33723614adba089e. Core d0ab5386 separately corrects a reviewed test-only port allocation race by binding both listeners before releasing either. Its source review is clear, SHA-256 adba1c3b548f639fc474029086756ec1f09fc55abb01f94834a0be8ca3742741. New CI and the full runtime retry remain pending; no configured realm operation has occurred.
• Storage recovery preserved 37 completed test-run directories in a verified persistent archive before removing their redundant temporary copies. Every file hash and original timestamp matched; all required build and predecessor inputs remain in place. The operation reclaimed 781,479,936 bytes, bringing temporary free space to 3,005,218,816 bytes. Archive SHA-256 2ec12dbdb9dff1f8b84e47e66228edff7521ed61d469ca26ec42971203f90d34; completed-operation audit SHA-256 2a3d765501b882da7db60beea4cbf82da59198f8e666403587a55d359558165e. All six proposed deduplication or alternative cleanup plans were superseded without execution. Core d0ab5386 passes all CI checks. The complete seven-scenario retry is starting at d0ab5386 and Collection 175c3ffc; no new companion result is claimed yet.
• The final seven-case retry passes at Core d0ab53864fca2b42b17429624c8cc496882ef19c and Collection 175c3ffc91956ddaac4c921e0ec66d0da65d0d71. The five order cases took 63.41, 7.43, 12.55, 14.04 and 9.68 seconds. The fixed companion route passed in 289.26 seconds and restart, Transfer and lost acknowledgement passed in 149.90 seconds. All capture, exact-case, source, Wasm, process cleanup and outer command checks pass. The retained Wasm SHA-256 is f4601db44f83e8894664e9b9cc55a0652264564eec3f9238380046a7fe54988b. Independent evidence review and the unchanged 10, 25 and 100 bot measurements are next.
• Independent review clears the complete seven-case runtime and raw behavioral observations, report SHA-256 354fb80f0ed2f62e6267c70904c33cc0e41db64ec3c322bf508853acdab25d13. A fresh, reviewed cleanup removed only 42 regenerable compiler files and recovered 398,544,896 bytes. Its completed result was independently checked. The unchanged persistent-load batch has started its fresh build at Core d0ab5386 and Collection 175c3ffc; no load count has passed yet at these revisions.
• Final persistent load passes at Core d0ab5386 and Collection 175c3ffc. Ten bots for 60 seconds produced 595 decisions and 585 completed movement legs, with maximum lag 5.135 ms. Twenty-five bots for 90 seconds produced 2,240 decisions and 2,215 legs, with maximum lag 5.408 ms. One hundred bots for 300 seconds produced 9,568 decisions and 9,536 legs, with maximum lag 2.513268 seconds. Every bot progressed and crossed the declared wall route correctly. All build, source, input, capture, process and cleanup checks pass. Independent review is clear, report SHA-256 a8b8c881e2d38f905c67a7b6cb3c9e1045ff68baa423e0be817802707801f5bb. These results cover the measured movement workload; imported-world gameplay acceptance remains pending.
• Core PR505 merged normally as 43592ceecf044ed6b6701bc42486ca6ab6ee981a after current CI, independent source review, all seven scenarios and the persistent load batch passed. Its tree is identical to the tested d0ab5386 tree. Collection PR35 is open at 69dd32afd9f1cbd080f6a72c92a0ed23874616f6. Its two-file CI addition preserves the tested playerbots source and binds all four candidate Core checkouts to the actual merge. The canonical PR workflow is running the 30 journeys, Warrior level-gap case, 44 lifecycle cases, both companions and existing durable and Transfer suites. A duplicate push-triggered workflow was cancelled; the complete PR-triggered workflow remains active.
• Canonical Collection run 34549049867 exposed a Mage staging observation race. One of ten Mage journeys completed; nine failed before ordinary journey activation because the test snapshot still held no prerequisite refusal timestamp. The Package deliberately leaves that field empty until the scheduled fault pass performs the real refusal. The test asserted the timestamp immediately after sequential reads. The failed artifact is saved, with matching source and case-list identities. The correction will wait within a fixed deadline for the genuine refusal before saving the staged snapshot. It will preserve the existing gameplay assertions and 600-second journey limit. Other canonical jobs continue; the 30-journey requirement has not passed.
• Canonical run 34549049867 is complete and retained as failed evidence. Warrior ten journeys plus the level-gap case pass; Priest ten journeys pass; Mage one journey passes and nine fail before activation. Composition passes 42 of 46 cases, including both companions. Existing durable tests pass 147 of 151. All six artifacts and nine job logs are retained. Independent CI review SHA-256 aea26f7ed027ad79f5149849a8e1b62da7caad1c7b25059386605656ddab8871. No complete canonical pass is claimed.
• The final cast correction starts and checks the real Waiting cast, sets the current objective deadline, then runs ordinary expiry and parks the bot in one fixture transaction. Independent source review passes at Core 077ea21f and Collection f014bd56, report SHA-256 a77acb21689830b5425e2577800912b550fab6381ca2d9bffd60a7d0c810cf27. The composed six-case run uses Core fc69bc76 and Collection f014bd56, content BLAKE3 77d6e34ea64be543b9815fb69f014a714da43f6f6d7a890f41622816785b2e15. Its reviewed plan SHA-256 is f1a694bda314388c23854c41f7eae214c76ce96afdc9636a7103ac931b7b85ab. Execution is starting; no pass is claimed.
• The six-case run at Core fc69bc76 and Collection f014bd56 is closed with five passes and one failure. All four corrected Quest recovery cases and Cohort Attack waiting pass. The cast helper succeeds, but the test expects an expired return-home objective to cancel an unrelated Recovery heal. The saved state shows that the current ownership rule correctly preserves the heal. Timing alone does not explain this failure. Every case ran once; source, Wasm, capture and process cleanup checks pass. Pre/post Wasm SHA-256 d4f2a0d8dfd3faa80fa7caab6cb9bbb8e8d86e6addce43320f29f56334d8f4da. Result report SHA-256 d91b4c9c3ac7c108eb6651ca9318cb394624ecf74b8a07b024f38da55a13245e.
• The obsolete Home-deadline cast assertion is replaced by a distinct ownership check. Its atomic fixture proves the same real Waiting Recovery cast survives the expired Home pass. The caller then requires exact normal cast resolution, a positive Core healing event, and deferral of the same Home objective when it resumes. Existing named Cast expiry coverage remains intact. Independent source review passes, report SHA-256 2c4cef4b3cb8abd7d3cba0ac822ac4399122f23e501a6493cceaa7d4c212cfa4. A single-case run is starting at Core 8a40bef8 and Collection 4f2739ae, content BLAKE3 5428cd7eae3cefb02189651d51421e57b8c0c1867194402d8ad5c771636c831c. The five prior passes remain recorded; full canonical CI remains mandatory.
• The revised tactical-cast case passes at Core 8a40bef8 and Collection 4f2739ae. An exact Waiting cast survives expired Home with no Deadline, resolves with 50 health restored, then an ordinary pass defers the same Home identity and deadline while preserving the resolved cast. All build, capture, source, Wasm and cleanup controls pass. Independent runtime review SHA-256 71f74431295a6102dd8334ba0e8c4867a804a99a7216cd40ed769a23c14f4d58. Final Core 78839ae0 removes two redundant locals to satisfy Clippy; behavior is unchanged.
• Core #509 merged normally as 67852c4ffb85794c89a5c1f515150ee8aca774d3 after all seven checks passed at reviewed head 78839ae0. The orchestrator verified that the reviewed head is an ancestor of the merge and both have tree f1958d0c80e8493efc3f98d5480b9b3c0eae8dad. Collection CI can retain its immutable Core pin. Package acceptance remains pending.
• The captured Transfer failure showed normal party reconciliation had already restored all four expected members while the exact arrival remained fenced. The fixture expected only the earlier three-member cache. Collection adb46d86 validates either exact state and stages the same missing-companion cache and loot-rule conflict atomically. Both source reviews pass; independent report SHA-256 110ae4696e6663cf4b61fc39d6468c702b671481161ce2fc0d144bbe484b8a8f. Production Transfer behavior is unchanged. Runtime acceptance is pending in canonical run 34558414639.
• Partial results from superseded run 34556764385 exposed two control-change caller defects. The delayed-command comparison crossed a legitimate active-to-inactive transition after leadership changed. The resurrection check read a latest-result field that a later scheduled RecordOnly pass could replace. Core #510 settles authority before comparing the delayed command and checks cancellation history for the exact accepted Resurrection Candidate. Both corrections have independent Standards and Spec review. The combined source retains all 44 lifecycle cases; runtime acceptance remains pending.
• Run 34558414639 passed all 27 Transfer cases at Collection adb46d86. The party-mirror caller captured the original three-member state, injected its fault, recovered through the real Gateway and retained the arrival until party repair. This proves that branch and the complete caller. The earlier failed run supplies the observed four-member precondition; the new normalization of that input has source review but no observed execution yet.
• The fixture now checks that the retained Home Objective identity matches the cast candidate. Test-process cleanup waits for live members of its owned process group through the existing grace period while keeping the leader unreaped. Owned subprocess checks passed for graceful completion, resistant-child cleanup and timeout. The final test measured 0.063808016 seconds between TERM and KILL against a configured test grace of 0.05 seconds. These changes and the final resurrection history check have passed review.
• The current Transfer job in run 34560742683 passed all 27 cases. Its party-mirror test directly exercised the already-restored four-member input, retained the exact arrival fence while staging the three-member fault, and proved real Gateway repair. The earlier corrected run passed the original three-member input. Current Transfer artifact ZIP SHA-256 is 43ed4e9fc87e8a7289432d4c6fe599ede3e97cdd3c2b54d510025b9060c93422.
• Current composition job 103142859665 completed 45 of 46 cases successfully. All 44 lifecycle cases and their execution checker passed, as did restart, Transfer and lost acknowledgement. The fixed companion route failed at its two-second controlled-target receipt comparison. Cast and physical receipt counts stayed 12 and 1; one 20-damage projectile impact arrived in the observation stream. The saved held and stable source snapshots both show target health 813 and that impact, while the separately queried held extra field omitted it. Production changes are not justified by this evidence. The test needs a settled projectile and observation boundary before exact stability comparison.
• Final combined run 34565025373 passed all 46 exact composition cases at Core 08ea73b2 and Package 67f325d. The execution record confirms every case passed with command status zero. The lifecycle checker accepted 15 actions, 75 outcome cells, 44 source cases and both reviewed unreachable contracts. The fixed companion route passed in 306.25 seconds. Its controlled-target boundary retained zero attack or cast handles, one unchanged control aura, exact cast and physical receipts through settling, then unchanged full receipts and health through the stable hold. Independent artifact review is clear. Composition archive SHA-256 cac093a9af56eab16285b65546fe6ae5371a87bc481ae2190440870a18f5cbd8; job log SHA-256 bd8d46c953ed988d2218660858caefe5a6dc96076de954d3b417d2175ddb570e.
• Completed run 34565025373 passed eight of nine jobs. All thirty exact journeys and the Warrior ordinary-kill level-gap case passed, with every required stage captured. Composition passed all 46 cases and Transfer all 27. Durable executed all 151 cases and failed only the provisioning-baseline assertion. Final evidence review SHA-256 548630544b9c03f7b21d2718770e6a66ba426ee6132e194db7764c1cc892505d. All review threads were resolved before the next source update.
• The two affected companion cases passed at Core 2347fca0 and unchanged Playerbots content from Package 67f325d. Explicit heal cancellation took 6.08 seconds; completion-time NoLineOfSight took 10.43 seconds. Both retained the exact cast result, no pending cast, productive Provisioning and physical Follow movement. Build and cleanup checks passed. The wrapper exited 1 because its Python import generated a bytecode cache in the Collection. Supplemental cleanup restored both sources clean, and the original nonzero result remains recorded. Root verified all 63 retained file hashes. Behavioral evidence review SHA-256 1e2b02c6e500cb871ff6ce7a005e28830616598e63efca72cb06447cffdc2a3e.
• Current run 34569550383 has reviewed passes for all 46 lifecycle and companion composition cases, all 27 Transfer cases, ten Mage journeys and ten Priest journeys. Every class journey retains staged, begun, first-decision and completed evidence. The fixed companion route keeps exact cast, impact and physical receipts, health, control Aura and empty owned Handles at its stable boundary. Warrior and the durable suite remain pending, so this is not a complete acceptance result.
• Final canonical run 34569550383 passed all nine jobs. Reviewed coverage is 30 journeys, one ordinary Warrior level-gap case, 151 durable cases, 44 lifecycle cases, both companion scenarios and 27 Transfer cases. Package #35 merged normally by rebase as a77013d13ca74c54a0a6f09e9f37e8e94dd31c7a at 07:30 UTC. Root verified full tree equality to the tested cb74 head and unchanged playerbots content. The duplicate run required one repeated module build to clear branch policy; its 1652 tests passed. PB-011 remains open only for criterion 8, measured imported gameplay capacity.
• Final canonical Collection run 34667391293 passed all nine jobs at Core d183fe5b589983ddfb3909450c759a42c25f40dd and Collection 915aec37a2a41511aeabbe125d6f4fa2f694a407. Reviewed results are 30 of 30 unique class journeys, all 13 catalog Quests rewarded in each, 156 durable cases, 46 composition cases, 27 Transfer cases and one ordinary Warrior level-gap case. Every navigation-fault journey recorded two approaches, a 30-second destination deferral, alternate Quest work, geometry restoration and resumed Quest 783. Core Rust, Runtime Scripts and durable CI also passed. Independent final audit SHA-256 b08a3ca26d31c210f12a2aa524e9075b279e73f092f406155116ab40717613e8; extracted evidence manifest a17c503b77b236d73b6e34657835ffee189df4e3a430224d1acb076d4fd7c591.
• The corrected imported hour completed at 09:52:22 UTC on 12 September. It retained 3,600 measured seconds for 25 bots, 89,921 decisions and 38,580 completed movement legs. Maximum scheduler lag was 514,616 microseconds. Every bot was processed and showed authoritative progress. Both archives finalized without errors, and private node shutdown and directory removal passed. Run SHA-256 04813955217885176ea3c6656eece0d106b8ef27b4d8b41fa602770ca664b544. The artifact retains its actual Core 85c2c204 and Collection e9309134 identities.
• The unchanged behavior verifier found zero refusal-spam candidates and 505 intervals needing causal assessment: 406 without movement, damage, XP or Quest progress, and 99 without XP or Quest advancement. The longest advancement gap was 2,860.161 seconds. Behavior report SHA-256 3ce0d4986119e2594b524f50a9c7b5db77da5076cf2e9d3e358b65f64ce87c77. Geometry assessed 112,823 segments and flagged six consecutive entity-position segments; all emitted splines were clear. Geometry report SHA-256 61b7678d5e9dc707e3ce14fda3704fe058807e0f3bc7fea0a39d9076e49d5978. The original failed findings remain preserved. Independent behavior and spline-lifecycle diagnosis are in progress; no imported acceptance is claimed.
• Independent replay of all processed snapshots in the 14 unfinished advancement gaps found 3,983 Quest 33 creature-target snapshots. Of these, 3,962 shared their target with another ungrouped bot. Each stalled bot selected 4 to 13 distinct targets and moved across a 94.1 to 186.0 yard coordinate envelope. The longest gap lasted 2,860.161 seconds. Source prefers the one catalog target before an active alternative and treats foreign-tagged creatures as valid autonomous work. The trace does not include private Loot Tag rows and cannot name each owner, but ungrouped attackers cannot all earn the same tag-owned reward. Root accepts this as a productive-target selection defect. Behavior acceptance remains failed. The correction must respect Loot Tag ownership for autonomous Quest and Grind work, preserve a useful active target, and leave Defense and human-directed assistance available. Final independent assessment SHA-256 7cc9da603b061132121ad2784c2b4e777e40b6333392ae6864439de2bdd6dc72.
• Independent geometry diagnosis reproduces all six flagged updates as exact Core f32 interpolation on clear retained splines. Both endpoints are walkable. The reviewed checker preserves ordinary segment checks and resolves a blocked rounded chord only with exact parent identity, simulation timestamps, reproduced endpoints and clear parent geometry. Its 16 focused tests pass, including refusal of off-path, reversed, missing-clock and blocked-parent cases. Verifier SHA-256 660c2cff67da372b0c9c560cc5ce672fc2139dbb3189f8846d1563100b33a85d. The full saved-hour assessment is clear: zero blocked paths, missing segments or unexplained discontinuities, with all six original findings retained alongside their exact proof. Result SHA-256 d7562c7de2f0ff5b5da065fb68ec96682cf0c95dd82abb31c73c48d67d736e80; diagnosis 72c40f3e7f265a747aaedcd8542c0e6977ee462a5071daa033142bf0f3e5588d. Root verified all six original identities and blocked cells remain recorded. The original v2 failed report remains unchanged. This accepts static geometry evidence only; behavior still fails and visible client motion remains pending.
• Canonical Collection run 34685881745 completed successfully at fe8b01b3082d00f0df29afa20d75cb9fb6c769ae. Root verified all nine jobs passed. Core CI also passes at 0a7e123e. These are the current remote sources before the productive-target correction; they do not resolve the failed imported behavior result. The final correction will receive its own complete CI and evidence review.
• A focused Quest case reproduces the productive-target defect at Core 45f6469156185d84e160eb7db0023ba0552df6c3 and Collection d16f7e0192205004ef8375b47653e80da95984ad. Production behavior is unchanged from the current PR sources. A second ungrouped Character creates the live Loot Tag through real positive threat. The subject selects that foreign-tagged target while an eligible alternative exists. Root accepts the saved state as causal failing evidence. Snapshot SHA-256 4ce20028c864a6271ca1ca36cd93c34abfdda9047423751b9ea8eded7af16e96; log 13084f0be7ef69a5032203425d7c4c7044a6110f82eedb7c121d1023196e2c43. The paired Grind case failed setup because normal spawn scatter invalidated an exact position assumption. Its original failure remains recorded. A reviewed caller correction derives target positions from the actual frozen subject and records the exact new target identities; only Grind is being repeated with the same Wasm.
• The productive-target correction passes all three focused cases at Core 8692be3d15cb0eda70ddd9d1cbd73486f7f3ea8c and Collection c46e9e3ec62434cca60eb8d69984d11bc918542e. Quest passes in 50.038 seconds: it avoids a foreign solo tag, retains an eligible alternative through primary respawn, switches when that alternative becomes foreign, and earns ordinary Quest 7 credit with sole corpse entitlement. Active Grind passes in 9.016 seconds and cancels its old foreground approach before switching. The Core Loot Tag fixture passes in 4.266 seconds, including party departure and revocation. Root and independent Standards and Spec review are clear. Result SHA-256 181f0b5d4059507c5799eb43402979c436f04fde1f88ca3cce893f4adca6c916; snapshot receipt f619b1c04ed9284fd840386b66530fc4d2c4b3ba138c330f857c9dbb8fdca431. Build, exact Wasm, source, inventory and private process cleanup passed. Both earlier causal failures remain preserved.
• A fresh imported-world run has started on the recorded corrected Wasm, SHA-256 3f49786c9c86591e8cef75a7d6e06dd13074b39b3b3eb45744ce5cb34564389e, 9,662,921 bytes. Build manifest SHA-256 256f4550a177f424693cd2d9228a3a4e38f7f62af2a7e93e4563559cf8c8b6ec. The 39 immutable Package input files match that manifest exactly. Reviewed run inputs SHA-256 3c9993dfd2198580edd9c779de108525c82f7a6ff944715163ee824c1336a563. The imported content, geometry, starting positions, 25-bot roster, 3,600-second measurement and behavior criteria are unchanged. The accepted geometry v3 checker remains in use. This is a run in progress, with no new imported acceptance result yet.
• The corrected imported hour at Core 8692be3d and Collection c46e9e3e completed all 3,600 measured seconds with 25 bots at 12:16:47 UTC on 12 September. Owned-node shutdown and private-directory removal passed. It processed 90,444 decisions over 7,181 scheduler passes with maximum scheduling lag 508,838 microseconds, zero lost due bots and zero reducer rollbacks. Peak Standalone RSS was 878,276,608 bytes; minimum free storage was 9,549,811,712 bytes. Compared with the prior full hour, bots at level 3 or higher increased from 9 to 20, Quest 33 rewards increased from 8 to 14, and unfinished advancement intervals fell from 14 to six. This run still fails acceptance. Behavior analysis identifies three RecoveryCapacity Refusal bursts and six unfinished advancement intervals. Static movement has zero blocked or incomplete segments, but two ghost relocations are real defects: a retained zero-duration stop spline returns the released ghost from its graveyard to the death position on the next tick. The original geometry result remains incomplete. Run SHA-256 2996e03e22bd47df40bfb3f7438b074b88888f9648ceb0cfa8ca4a69c4ec7e11; behavior d6635be9d88cb5a5d79da548f8761062c692008efc0c4e72a1a184668086b80b; geometry c195f098f886110d40517c6b46c151eb603c1a60453697f9756384747f2888da.
• The final 25-bot observation completed 3,600 seconds with 7,180 passes and 90,466 decisions. Every bot was processed. Maximum scheduler lag was 505,859 microseconds; all 27 deferred observations were accounted for. Peak process memory was 885,452,800 bytes. The run still fails imported behavior acceptance, so these measurements do not yet close the deployment-capacity criterion. Original run SHA-256 c53438ee525ccb3d879e1b5d92189d4763505d633ca46dc8a383585b379482f4.
• The next capacity observation is running with the combined corrections at Core 916bac3a and Collection 4e8d98e3. It retains 25 level-one bots, 3,600 measured seconds, automatic provisioning, ordinary XP and Quest credit, and the same imported content and geometry. Final Package CI runs alongside it. Capacity acceptance remains open until capture and independent behavior assessment pass.
• The latest imported hour at Core 916bac3a and Collection 4e8d98e3 closed successfully at 22:00:25 UTC on 12 September. It recorded 7,176 measured passes, 90,601 decisions, maximum scheduler lag 507,060 microseconds and no lost due bots. Eighteen bots ended at level 4 and seven at level 5. Peak RSS was 869,617,664 bytes; process CPU was 896.13 seconds. Private cleanup passed. Run SHA-256 8e8814d01896161826735d155abad3967bf786460733c00110c88fbd37a5a49c. Three confirmed Recovery self-heal Refusal bursts keep behavior and capacity acceptance open.
• The corrected capacity observation began preparation at 23:35 UTC on 12 September, using the recorded Core 27d6f533 and Collection 54a7d39d build. It retains 25 level-one bots, 3,600 measured seconds, ordinary progression and the same resource and behavior criteria. Input manifest SHA-256 4e4361e29ed53ba6486bbd0f550177810199fd7d15c25907d7bcba36948c6c0b. Capture and independent assessment remain pending.

Current notes

• All nine criteria are verified and final source PRs are merged. The accepted imported hour establishes capacity for 25 bots in its measured setup.
• All 25 level-one bots earned the six starter Quest rewards and ended at level 4 or 5. The attended real-client session remains a PB012 requirement.

Complete the additive state cutover with one durable controller selector, compare record-only decisions, activate a small cohort, gather imported-world and attended client evidence, then delete the displaced behavior source, obsolete tests, and active old state path. Leave old table retirement as its separately reviewed migration when required by repository rules.

Acceptance criteria

• The additive state migration defines defaults and backfill, and exactly one durable selector makes one runner eligible to control each bot.
• Record-only comparison on fixed facts issues no gameplay actions and reports deterministic policy disagreements before cohort activation.
• A small active cohort can freeze without a second controller and can resume after the failure reason is recorded.
• Both automated release suites from PB-011 pass again with the new runner selected and no old runner actions.
• A one-hour imported-world observation runs 25 bots and records no unexplained stall, repeated Refusal spam, visible clipping, or idle injured healer.
• An attended session in a real unmodified 1.12.1 client exercises the companion route and signs off follow, stay, assist, target choice, crowd-control respect, real casts, death, regrouping, and instance entry and exit.
• If the implementation environment cannot run the real client, the exact observation script, fixture revisions, capture format, and diagnostic reads are ready, and this acceptance item remains explicitly pending human evidence.
• The displaced behavior source, obsolete helper-shape tests, broad internal exports, active old controller path, and unused old state access are deleted after acceptance.
• Repository checks prove the deleted behavior source cannot regain control and no production code reads or writes obsolete active state.
• Any physical old-table removal has its own reviewed migration plan and does not weaken the completed single-controller cutover.

Verified criteria: 7/10.

The additive state migration defines defaults and backfill, and exactly one durable selector makes one runner eligible to control each bot.

Reviewed by Orchestrator and independent criterion review.

• The accepted five-case transition run at Core d201581f and Collection a19229f8 proves populated migration, bounded batches, restart, replay and Transfer ownership. All command and cleanup results pass. Root acceptance SHA-256 b34ff18ff07549633e250b20100444d023604eb58d652245ddd870dc2651df8a. The final source review confirms one durable selector and preserved Legacy migration defaults.

Record-only comparison on fixed facts issues no gameplay actions and reports deterministic policy disagreements before cohort activation.

Reviewed by Orchestrator and independent final evidence review.

• The Priest comparison case passed in both composition and durable execution in canonical run 34667391293 at Core d183fe5b and Collection 915aec37. Four passes over fixed entity, party and personality facts recorded the same new Heal decision and the source-derived Legacy Follow disagreement before Cohort activation. The complete captured gameplay state stayed unchanged, with no Action, pending cast, cast event, melee or moving spline. Independent criterion audit SHA-256 682f366b7f09d77a1a55bd68d1488f985edea39224616318511aef07407e58da.

A small active cohort can freeze without a second controller and can resume after the failure reason is recorded.

Reviewed by Orchestrator and independent final evidence review.

• The bounded-failure-memory Runner case passed in the same 156-case durable job. After recorded NoMovement and destination deferral, the sole Bot remained Frozen with exact Runner and physical state for six seconds. No foreground Action, pending cast or moving spline appeared. Cohort then resumed with a later observation and the same Objective, failure and deferral. Independent criterion audit SHA-256 682f366b7f09d77a1a55bd68d1488f985edea39224616318511aef07407e58da.

Both automated release suites from PB-011 pass again with the new runner selected and no old runner actions.

Reviewed by Orchestrator and independent final evidence review.

• All 30 class journeys and all 46 lifecycle and companion cases passed in canonical run 34667391293. Every journey-begun and completed snapshot selected Cohort. The companion route retained Cohort through each landing and empty prior-location action state. The exclusive dispatcher and retained outcome captures exclude Legacy actions from these runs. Independent criterion audit SHA-256 682f366b7f09d77a1a55bd68d1488f985edea39224616318511aef07407e58da.

If the implementation environment cannot run the real client, the exact observation script, fixture revisions, capture format, and diagnostic reads are ready, and this acceptance item remains explicitly pending human evidence.

Reviewed by Orchestrator, source and procedure review.

• The guide retains all command blocks, nine attended steps and actual gameplay evidence identities. It now records the supplied Operator result and resumes at Service Reconciliation, then Gateway replacement, Map 36 vmap import and staging. Guide SHA-256 7e141feeb7c6a056831dfb9b66b52a9a687e8f3045d1d0bec5f0de2f8b6e5f97; unchanged template 24852e8fa736cbcef51bbf0e96888bbe964a18600d8b6df1f3db080901ff992b. This accepts handoff preparation only. Running candidate bindings, session GUIDs and human observations remain pending.

Any physical old-table removal has its own reviewed migration plan and does not weaken the completed single-controller cutover.

Reviewed by Orchestrator and independent Standards and Spec review.

• No physical table removal ships with this release. The reviewed plan preserves pkg_playerbots_goal, its Character-delete and not-transported lifecycle, and the indexed kind-5 predecessor Transfer check. Controller::Legacy, its serialized order and the roster default remain separate compatibility requirements. Current Cohort personality rows are retained.
• Logical cutover requires zero Legacy controllers and no unsettled bot Transfer Intent, source Escrow, Runner checkpoint or kind-5 goal row across the applicable Shards. Inactive old goal rows may remain. Physical removal additionally requires an inventory of every old row, a separate bounded and replay-safe Operator cleanup, and stable zero counts. Unknown goal kinds stop that cleanup for review.
• A later schema change must remove the old table and its compatibility readers together, regenerate complete Gateway bindings, pass schema parity and preflight, and receive the required human schema review and named-Realm deployment approval. Its recovery procedure cannot assume that a Package which reads the dropped table remains usable. No Realm operation or physical deletion has occurred.
• The plan binds prepared Core ec578c40 and Collection 5ae91707 while preserving the original inventory. Full plan: playerbots-table-retirement.md, SHA-256 794f395598809fc3035c1add8393f3f1d1f2330fa332e71ee2fcb9286647dd8a. Original criterion review 6d04b347a12ac37f6b2cd0afa3ef1dddfbdcfd4b8603c47eb6ac0f3760a43ab9; current Collection compatibility review 3c8127ec7089682fc36311fd84eb45db46c5ad85deba2a6b971e92157cf00c90. All compatibility owners remain unchanged.

A one-hour imported-world observation runs 25 bots and records no unexplained stall, repeated Refusal spam, visible clipping, or idle injured healer.

Reviewed by Orchestrator and independent complete-span, geometry and capacity reviews.

• The accepted v19 observation ran 25 bots for 3,600 seconds on the pinned imported world. All 90,583 timings are present. The 269 flagged intervals have exact work, Objective and target accounting. There is no unexplained stall, repeated Refusal burst or idle injured healer.
• Static Core paths are clear across 146,375 segments, including 21 rounded samples proved by exact clear parent splines. Rendered-client evidence remains separate. Three Grind intervals remain right-censored; their 70 cancellations are accounted for by 61 foreign targets, five dead targets, two outside-scope targets and two immediate available-target reassignments.
• Causal review SHA-256 5f188ab6859c143770c01060f9b995a6b7b996f3eabd63efa58b11af1c202fa4; target and death supplement 98a9e31ebba7b97800e1badf69592a35a40af2b6e13ef9fce12d8c3a630abca5; root acceptance e681419b347641418f4194c55f756ded100449405aa3c6a2d41a5d0b856deaa9.

Imported world observation

• observer: Orchestrator with independent capacity and complete-span review
• observed at: 2026-09-13T02:30:51Z
• core revision: 27d6f5337276dc461cdb376927ada8f4e5c9017b
• package revision: 54a7d39dd869e0aaf81c707e4b30728bf008d536
• content revision: 7e2e5ca12108505cbce8863a9d0d2d592ba3c52b8f53c0cadb0a5767b8f9450b
• geometry revision: 3451e2610b6c45176c815d34841e485c1ef5e124a21831582919d8758e5152dc
• minutes: 60
• bots: 25
• evidence: ['Captured run SHA-256 5fde8f19391b154f66e43135cbfae75aed5eea79a3f2a4c0e29e295beaf8d463.', 'Accepted raw capture, three frozen analyses and independent reviews are bound by root receipt e681419b347641418f4194c55f756ded100449405aa3c6a2d41a5d0b856deaa9.']

Attended client observation

Pending. Required before completion.

Execution model: gpt-6-astra, xhigh effort. Repositories: Official Package Collection controller cutover and source retirement, LyraCore additive state migration and architecture checks, Imported-world observation and attended client evidence.

Recorded evidence

• Read-only filesystem discovery found a local executable whose embedded FileVersion is 1.12.1.5875, plus the client MPQ archives. Executable SHA-256 b4756d38ef207c02ed651f4952bd89a70b4857b73a33413339e1b285b28d2dc7. Extracted geometry is not yet prepared. A local package patch archive is present, so an unmodified-client acceptance claim remains unproved. No client was launched and the configured dev realm was not changed.
• Read-only local input discovery parsed 22,357 spells, 432 talents, a 25-cell Northshire terrain sample, and placed world collision geometry. Six M2 models failed parsing. The existing importer binary has unverified source provenance, so this proves available inputs only. No database was requested, no rows were imported and no client was launched. The final pinned importer, complete coverage, one-hour imported run and attended-client evidence remain required.
• A pinned offline alliance-eastern import plan passed in 18.17 seconds with 8587 creature placements and no EventAI compatibility findings. Every supported quest template, starter, ender and objective is present. Four later quests require level 7 in the actual input, so imported-world acceptance must earn that level through ordinary gameplay. No rows were applied and no Shard was contacted. Full route geometry, the one-hour run and attended-client evidence remain pending.
• Pinned read-only terrain extraction passed all three alliance-eastern slice self-checks and derived 28,556 cells from 142 tile reads. Separate DBC, spell and talent preflights passed with 432 AreaTriggers, 22,357 spells and 432 talents. No Shard was contacted and no rows were applied. These checks establish available inputs; navigation, collision coverage and gameplay acceptance remain pending.
• The pinned importer completed the full offline alliance-eastern navigation plan in 129.79 seconds. Human Start, Dun Morogh Start and Loch Modan Start self-checks passed. The plan contains 23,579 unique navigation rows in 1,010 batches, with peak memory 203,912 KiB. Nine source archives were reverified. This dry run applied no rows and does not prove imported gameplay or final line-of-sight coverage.
• Independent Standards and Spec source review passes at Core0c167f61c5cda723ec963dd9917d8efdd84ed7dc and Package605129fbde774acbfc97d51cf2e01353c27ad1f0. The review report SHA-256 is0070d55a0a9e966142379b8046ca286ab4a4e7a09135a6fef696f1692f16d8c6. Accepted predecessor identity remains an execution-time input, and every runtime and release observation remains pending.
• The reviewed transition patches now compose onto the provisional acceptance pair as Core4065f73e and Package82acfbeb. Core patches remain identical; the sole Package conflict retains both companion and controller-transition fixture registrations. Composition review passes. The accepted predecessor manifest remains a mandatory future input; no transition runtime has run.
• The transition patches compose onto the current acceptance base as Core 145c8573388e416540fe9f312a0eb25ee1c5422c and Package 2f3da0fe6d0adb3f0498057ff623063d0920ea0b. Independent review confirms all three Core and four Package patches are unchanged. The accepted PB-011 predecessor Wasm, manifest and digest remain required future inputs. No transition runtime has run.
• Independent scheduling review found a dependency cycle between PB-011 measured gameplay capacity and the PB-012 imported hour. Moving the PB-011 dependency to completion permits the shared observation after the other PB-011 criteria pass. Review SHA-256 43b47346a3fe590dbec2a0961cf5bcde031d5ed13b4b6de602e1138bfd64ee79.
• The imported-hour checker now accepts the exact lowercase hex byte arrays emitted by the pinned CLI, including empty obstruction values. Actual retained navigation output confirmed the format. Verifier SHA-256 99a2799f6d92aceca87ab55f4dd0861228d5901bbcbf59a8e45d563ed1d899f5 passed all 12 focused tests and independent Standards and Spec review, SHA-256 e012bf08a208d3758da62d1e797368eb62b0461679a95b63acbd58c017577d54. The revised readiness plan is 13597de60aa7b3f477253cd5ac3176817d9a6e20b89d30afb329cfeb01cf21c2. These are preparation checks; runtime acceptance remains pending.
• Core #513 composition passed independent Standards and Spec review. Its first CI found one Clippy failure: the populated migration test exceeded the 200-line limit. Commit 34be4d95 extracts four identical state snapshots into one helper while preserving all queries, boundaries and assertions. Independent review and focused local Clippy pass; the latter took 44.65 seconds. Final required CI is running at the corrected source. No transition runtime has run.
• Draft Core #513 has all seven current checks passing at 34be4d958ef92d8c34e9bef247ecb82cd87dbc61, tree feed26572b03c2ae933354533139c937954484d9. The durable main suite passed in 15 minutes 23 seconds, followed by auction and Gateway regressions. Independent composition and snapshot-extraction reviews are clear. CodeRabbit skipped automatic source review while the PR is draft. The five private transition cases remain required before readiness and merge. CI monitor SHA-256 25546746502614f5d9b415e82beccf6475bf1718a8e8b48f3ddb3fa38d526274.
• The fresh predecessor build passed at accepted Core 20524be6 and Collection a77013d1. All ten status rows and every final source, control, artifact and cleanup check passed. Root independently verified all 38 files across the four Package inventories, both clean source trees and restored Package entries. Wasm SHA-256 faea46ed842a110b2327892382c0119c2a92578a52d252169ba3f0d1a764e7b3, 9652044 bytes. Build-manifest SHA-256 86e534254686342da964779a188ee7b706e9dcf3fc39da885664365cb311d364. This is the accepted predecessor for the five transition cases; no transition case has run yet.
• The final Package composition is clean at 206ba3557d62b92018cafad2a947e845dfa80a78, tree 4b3ae8e8767de8f19672c5032d0291d4b4f085ce. All four reviewed transition edits match their original changed lines. A separate CI commit pins four workflows to reviewed Core 34be4d95. Formatting, workflow syntax, source ancestry and diff checks pass. Draft Collection #36 is open, and canonical release run 34575890235 has started. Composition report SHA-256 8b03e623722d67a039bbf2eea1e3fb8b98bd6173cd415742601574f5c57b9071. Actual transition acceptance remains pending.
• Independent Standards and Spec review clears final Package 206ba355. It confirms exact transition edits, retained PB-011 fixes, four Core pins, stored Legacy compatibility, Refusal before state changes, bounded sorted migration and Transfer ownership fences. Review SHA-256 2241f436708115648d0c48772d31ebf636bdd9a6c88d88f614fc8635971430f2. Runtime acceptance remains pending.
• The first final transition run completed all five cases at Core 34be4d95 and Collection 206ba355. New defaults, the populated 17-bot upgrade with restart and replay, and stale IN_TRANSIT handling passed. The two Escrow cases failed before begin_transfer ran because the CLI parsed a negative destination coordinate as an option. Overall status is failed. All artifact, source, process cleanup and Package restoration checks passed; candidate and post-runtime Wasm match. Result SHA-256 dd0947da797005cfa999503614afbb1e9e8b89d4d9b251877d739a64ff996542. The failed evidence remains unchanged.
• All five corrected migration cases passed at Core 0b277102 and Collection 8e3c913b. The run proved default Cohort selection, atomic Legacy selection Refusal, the populated 17-bot upgrade, bounded migration, restart, replay and both Transfer ownership boundaries. All final controls passed. Root verified the five exact results, logs, unchanged post-runtime Wasm and all 39 Package inventory files. Wasm SHA-256 84340fd6e1c373d661b4fe68b62b683687a9a7ed4253508c12d101ed7abd2dfb; manifest c3dbc849441c32b047b64f168dbd484383832b97e9aa6fcdc1fd36ec6f614e32. Root acceptance receipt 00c78de8c61b85d0a8a25306fbda42fe368021b89373a6f2ae66fc8f01183baa. Core #513 is ready for review; full release acceptance remains pending.
• All five migration cases passed again with the role-fixture correction at Core 0b277102 and Collection 1d87c29e. Root verified every case, process log, final control and all 39 Package inventory files. Wasm SHA-256 4d495a6fffad26535a6d3133bfdc2f58c8ed8714f97152d98ffae4e7a87c6b03, 9866826 bytes. Manifest SHA-256 ef4f9d0366019daa47aebadd218141abd35c2700c7d0c99f8918fbc3f9a48752. Root acceptance receipt 8ce9e05e9649c919f30c6601f2658de4b35c6084895a9153a90f533f1404af7f. Final full release run 34577820134 tests that exact Collection and Core pair.
• The first imported-world attempt passed its immediate source, executable, archive and resource checks, published the private Module, and imported DBC, spells and talents. World import then stopped at the 25-definition EventAI relay catalogue: the CLI received a multiline String without JSON encoding and returned HTTP 400 before the reducer ran. The observer and timed hour never started. The private process stopped and its fixture directory was removed. Both analyzer invocations stopped on missing capture input and produced no analysis report. Failed attempt result SHA-256 f0571360d52d6311b3a375a80578403689dccbcb890216efbfa56e4ca5de0bc4; closure report a0fc5f0685e96dc463febd3a2e8d3c1ed9fcc34734bbfa9a00902cf68bdf939f. No configured Realm was contacted.
• The stronger migration preservation assertions passed all five actual private cases at Core f3ea85d1 and Collection 1d87c29e. Final run, capture, cleanup, source and control checks passed. Combined Core #513 at 86653687 also includes the command-capacity caller correction. Collection #36 now uses 720d1c7, with unchanged playerbots content and four CI references updated to that Core. Complete canonical run 34580465137 is in progress. Earlier complete and failed runs remain retained.
• Core #514 contains the importer repair at 274b0194. Both affected String call sites use JSON encoding; numeric, Boolean and vector arguments keep their existing dispatcher. The reviewed private replay checks the original 25 Relay Definitions and a two-document Package Delta import before any hour-long retry. Local importer checks and replay remain pending.
• The exact command-capacity case passed once in 131.34 seconds at Core 86653687 and Collection 1d87c29e. Core-only Clippy, compilation, exact case listing, artifact capture and source restoration all passed. Root verified the case result, all four process logs and Wasm af301c014585ec5a749c3c3a5701b227b401f90d348e882609317759f4b0e6eb. The first local check stopped before runtime because it incorrectly included unrelated Package lints; that failure remains retained.
• Collection 1d87c29e composition passed 44 of 46 cases. The remaining two Runner callers expected same-value Cohort selection to initialize a bot parked by their fixture. Core 8cc71bf9 adds explicit Frozen staging in nine automatic Runner scenarios that shared this assumption. Manual stepping, controller idempotency and exact history-generation tests remain unchanged. Independent review is clear; the nine focused cases are being prepared. No Package gameplay code changed.
• The latest complete release run is 34582130469 on Collection aa03ec1d, tree 912cc999cec8000d6779f333fa25a06cd496d143, and Core 8cc71bf9, tree 062fd1663cc1c853e6a49e2518b2f5d606ef4d60. Only CI references changed in this Collection revision. Current PR #36 remains at 720d1c7 while its complete run closes, preserving its evidence. That run passed all 27 Transfer cases. Every earlier active full run continues to closure.
• The older complete run 34577163722 passed all 30 class journeys but failed its separate ordinary level-gap case after 600 seconds. The Warrior had eight kills and 608 XP, remained level 5, and retained Defense movement. This is under source and trace diagnosis against earlier passing runs. Its older revision alone does not explain the failure, and no correction is claimed yet. The latest full run remains in progress.
• Importer 274b0194 passed 367 tests with eight ignored, Clippy and release build. Binary SHA-256 f2fa5d64d6ca96c9f760831c510fce49775583b439b348a84f16344d2c25d8d4. Its private replay passed publish, DBC, spells and talents, then stopped at EventAI definition batch zero with HTTP 530, boolean must be 0 or 1: true. The original multiline parsing failure no longer occurred. The final 25-row assertion and two-document Package Delta replay did not run. Root verified all 29 captured files and eight build records. Process and fixture cleanup passed. Closure SHA-256 a0444f52a3d8a0761a5321a10677c2bb8a00e7f06ac6f387e1e47ef982ae0995. This is failed setup evidence, with no measured gameplay.
• Run 34577820134 is complete. All 30 class journeys and all 27 Transfer cases passed. The separate Warrior level-gap case passed in 448.56 seconds. Composition passed 44 of 46 cases. Durable execution stopped after the first Gateway target passed five of six cases; its remaining 145 Module cases did not run. These known caller failures have reviewed corrections. The separate earlier leveling failure remains under diagnosis. Final report SHA-256 443bd88c72ae8566f1c7dc1abbfe7dca8191522c67512d974f33c6f358e7377f.
• All nine corrected Runner activation cases passed at Core 8cc71bf9 and Collection 1d87c29e. Root verified each exact case, all process logs, 27 runtime snapshots, source restoration and cleanup. Wasm SHA-256 e7bbd760ecc1de6c2960a02e2a4901d4227c67d89255beb352310de2c05630ad. Acceptance receipt SHA-256 09a468e756adcf492694fe4339595828cc6bcd51a3b971a216c4416da39b73ae. This proves the nine callers; complete release acceptance remains pending.
• The failed leveling trace proves a post-resurrection liveness defect. Release Spirit moved the bot about 7460 yards, but the Runner retained its live pre-death Defense target. Each movement leg advanced, so no-progress recovery could not break the chase. Defense continued to outrank the retained ReturnHome Objective. The correction clears the transient Defense target and paired health observation on successful Resurrection while preserving Objective and Quest state. A deterministic death/release/resurrection regression is prepared. Earlier passing runs avoided this path and do not disprove it.
• The same trace exposed a separate level-gap fixture gap. Its XP sources and navigation were near 1200,1200, but its graveyard remained roughly 7460 yards away. The corrected fixture declares a local graveyard using the existing full-journey pattern. The test checks its identity, map, coordinates and retained placement. Enemy damage, XP, the 600-second bound and all existing progression assertions remain unchanged. Actual corrected runtime is pending.
• Importer c4d1db17 changes all seven Boolean fields in the six affected EventAI movement encodings to the Module format, 0 or 1. The actual batch regression failed on the old encoder and passed after correction; all 367 importer tests and Clippy pass. An earlier verification attempt reused the old executable across worktrees and is retained as a control failure. The corrected wrapper clears the affected package between revisions. Release build and a fresh private CLI replay continue.
• The corrected importer completed all 12 private replay steps. Actual SQL results contain 25 distinct Relay Definitions and spell 133 with cooldown 1500 ms and GCD 1000 ms after two Package Delta documents. Root verified all 37 captured files, native build records, exact source identities and process cleanup. Importer SHA-256 2aa6f100b0432fbeec71e2659686453a45d1e753840f9fad7f5c1e84a001b65c. Replay closure c3cf8539ed36af024c316d1f72483e457322e939de85f15772af1268b0466bca; root acceptance d650fa93f122194d38e4a18b31619a1ca39f3c10037b77611156b010c5fee9df. No measured gameplay ran.
• Independent Standards and Spec review clears the resurrection correction at Core d9b0617b and Collection 795524dd. The new fixture applies actual damage and parks the bot in one transaction, closing the scheduler gap before observation. The test checks full Objective and Quest retention, the same live attacker, both resurrection phases and the dead-hit no-op. Review SHA-256 e4fde796ec203d83ac4006ac371b32ce8352f4ae9e22dbf421fd81b9bad7b39e. The reviewed failing-old-code and corrected-code run has started; runtime acceptance remains pending.
• Run 34582130469 composition passed 45 of 46 cases, including both corrected Runner callers. The remaining command case received the exact Unchanged wire reply and retained a Finished Unchanged intent, but its later snapshot found no addon event row. That row expires after one second. Core 08a25f0d checks the already-captured exact client reply and retains all durable intent, receipt, order, Objective, inventory, Quest and party assertions. Source review passes. The full corrected runtime remains required.
• Core #513 now tests composition 6a96111f, tree b82fcfbc23f1128e8d04b71349021d6c52740fff, containing the resurrection test, reply observation correction and importer repair. Collection 796b978, tree 6586691a52ac465c9d586fb4eb5c487b9dad95bb, is prepared locally with four CI references to that Core. It has not started a new complete run. Existing full runs continue to closure. The earlier 720d1c7 durable run passed 141 of 151 cases; its two failures outside the corrected Runner callers are under diagnosis before final composition.
• Importer #514 merged normally at b85d99fcd30ab6313298ac034527129593a049e2 after all seven checks passed. Its full tree equals accepted importer c4d1db17. The private replay and build evidence remain unchanged.
• The resurrection old-code case reproduced the exact stale Defense failure. Corrected code cleared both transient attacker fields through Release Spirit and Spirit resurrection and preserved the complete Objective and Quest state. Attempts v2 and v3 then stopped at the next travel assertion because Provisioning consumed that pass. The v3 setup called provisioning while Frozen, where reconciliation intentionally does no work. Source diagnosis requires provisioning after atomic Cohort selection and parking. The level-gap check has not run. Both failed attempts and successful cleanup are retained.
• Core e877db92 adds explicit Frozen-to-Cohort setup to the automatic companion and provisioning callers. It also retains the exact wire reply correction. Source review passes; complete corrected runtime remains pending. Run 34582130469 has passed all ten Mage and all ten Priest journeys; durable and Warrior jobs continue. Older run 34580465137 is closed and retained with its known failures.
• Full run 34582130469 is closed. All 27 Transfer cases, all 30 journeys and the ordinary Warrior level-gap case pass. The run remains failed because composition passed 45 of 46 and durable passed 148 of 151. Final evidence report SHA-256 8b1092a1a4d6b4398ea1991e1d1c1b7af6df02351e7da6bbbf467e48b5435c7c. The issuer-fence caller drops only an assertion on immediately active state that the scheduled Runner may clear before a separate read; exact issuer history, fences, later inactive state, whole-row preservation and the Superseded receipt remain.
• Recovery setup v5 has independent Standards and Spec review at Core d18e185d. The 22-action Priest profile finishes within the bounded loop over single ordinary provisioning steps. The test observes its completed durable cycle and checks full-row preservation during recovery. Review SHA-256 a71440dc71cd76ff7c9a44ae2a077cf64a746b677da28b99a80463aa935a7842. Runtime is in progress; no new acceptance is recorded.
• Recovery validation v6 passed at Core ad83247e and Collection 795524dd. Root verified all 71 retained files. The old-code regression failed at the required stale Defense assertion; corrected recovery passed in 7.56 seconds and ordinary level-gap acceptance passed in 424.17 seconds. Every capture, Wasm, source and cleanup status passed. Result report SHA-256 dbc098960182bd63d42a23ab16043340c512d8b2316de155fd765fc3037222f9. Root acceptance SHA-256 d0e056a8117778e2bd6ee92408bcb6b78e6174e8b881dacdd63945144a0f5248. This proves command resumption and ordinary progression; imported geometry and attended client evidence remain separate.
• The final combined candidate is Core d201581fe099343a03e5fd970804ed53c4f8426c, tree 69f64e80f9ad1a7914a606956e42df4c9d849cfb, and Collection a19229f8647566327e1743b6be2878166aca8e8b, tree 80152e59cdba5461d24a9bca8d301ef1dd199764. Its playerbots content remains 64b7588d624480c7a4ce30500dbd33cfdd0dd06faafd3c04f9bb396473cff9bf. Canonical release CI 34590204967 and the reviewed five-case migration producer are active. No complete release acceptance is claimed.
• The final five-case migration producer passed at Core d201581f and Collection a19229f8. Root verified all 70 retained files, every command and capture status, all ten scoped cleanup commands, the exact four-Package inventory and identical candidate/post-runtime Wasm. Candidate Wasm SHA-256 122d2f9afd0d3a0a18329ec54f198a0af8d5f5b6bf0ef802e5bef0ce832e06d3, 9,688,931 bytes. Build manifest SHA-256 f09a5145ca69af00c4f2d120a9ccf00c8c36b3d89791ce046231e231081e961c. Independent root acceptance SHA-256 b34ff18ff07549633e250b20100444d023604eb58d652245ddd870dc2651df8a.
• Final hour input review is Standards CLEAR and Spec CLEAR. All 27 execution fields are resolved from the accepted migration artifact, all 39 Package files match the build inventory, and fixed input, resource and coverage controls remain unchanged. Review SHA-256 ffadc0c7d53b4c71702c3d859d99be26c5772780dc662200366484716b22870a. Input SHA-256 98aeac4e296d813f77bc078835bcb93e10ca48a1db6971f9267c36635d31cd89. This is preparation; the measured run has not started.
• The imported-world run started with input SHA-256 98aeac4e296d813f77bc078835bcb93e10ca48a1db6971f9267c36635d31cd89. Immediate execution preflight SHA-256 006b7e7086348c9ab86ad1938fc8f4b13a167e4c62f78d01502447c832753eab verified nine archives totaling 3,666,041,816 bytes, all fixed inputs and Package inventory, clean source revisions, no competing private runtime and the declared storage reserves. Data import is active; measurement and both independent analyses remain pending.
• Deadmines importer db150e90, tree 382ba7cf4a16a0343eb3569920de65f6c4ad29d2, passed native tests, Clippy and release build. Root verified all 13 command log identities and cleanup. Native acceptance SHA-256 eed718532f1b9d6789fd60b3253882cd250466d7b9ece99f9ae31812a8f78d04. Actual archive extraction failed at the MPHD flag assumption before geometry output. Failed result SHA-256 965cc195dc22a626a6cb42048bc57c77d39df518bb66e40e2ca4aca20b08859a remains retained with exact inputs and clean process closure.
• Imported-world run d201581f/a19229f8 closed as failed at 2026-09-11 11:23 UTC. All imports passed; playerbots_imported_stage refused bot 19 because its start lacked a walkable floor. No observer ran. Owned-node shutdown and private-directory cleanup passed. Both pinned analyzers returned 1 because observation files were absent. Root result SHA-256 39cc2f46273720de7fa4208166d2e98c3b2de0b889b396d41210f8bd9124bc54.
• Bounded Deadmines WDT inventory passed six parser checks and all four execution steps. WDT size 32,836 bytes, MVER 18, MPHD all zero, MAIN 36 active tile entries, MWMO empty, MODF absent. Inventory SHA-256 9ff735caf1b2118cc3898c0d8e7383a83065d84410a0c22184d70ba18fc78eb4. Result SHA-256 d0cbc6847d4d27dd247093e0cf95c4ea668202be8a0525ef64e7ebfcb0a46b5e. Source and process closure passed. This establishes the extraction format, not movement acceptance.
• Canonical run 34590204967 evidence review is complete. All 30 autonomous journeys and the ordinary Warrior level gap pass, as do 46 composition cases and 27 Transfer cases. Module tests pass 1652 cases. Durable tests pass 151 of 152. Final review SHA-256 89c19ca3f938f1bfb3f065ada3bc1ee8f707a5da5440b3ac937a80905e95a4ee.
• Provisioning correction de6ca5f0, tree d1c80afcf836008f5553b4d25226ebf064882e48, changes nine lines in the test. RED reproduces the exact failure when the explicit pass is 199,853 microseconds before repair eligibility. GREEN passes in 7.50 seconds with the same Wasm. Independent source review is clear. Root acceptance SHA-256 36180ed407ce5a371dbb07018937e6211a02825742a61309cfad1b0a5c3b83f0.
• Imported placement candidate Core 19ad08181dc2d6cd77e797616947c8cf9bb33f4c and Collection 6598bd0e6d0032905f0d02890ca81edba31a7755 retain level-one XP, class roles and geometry checks. The 100-candidate bound and sparse-search rollback cases passed independent source review, SHA-256 b05f86ea27040ff67f6918342df8485ef5e3eb4686a08f0c86d178a72242d87c. Focused runtime checks are in progress; this is not imported-world acceptance.
• Imported placement focused acceptance is complete. Root verified all 42 files in the final evidence inventory. The exact old-code blocked-floor failure and both corrected cases passed independent review. Root acceptance SHA-256 4741d60e2c6577f8881b965cd0c011463b984a2771771501bb1e8adbf0aad912. Independent review SHA-256 82c643cf3d693e369d238e7cedbb2819458421a5e88e489af4bb86ccc2e89d80. The first recording failure remains retained and supplies no corrected-code test result.
• Core 86e660f4 and Collection 50e2cd10 are the current PR heads. Core rebasing preserved the entire reviewed source tree. Its final test-only change combines the same two provisioning assertions in one loop to satisfy the function-length lint. Independent review and focused Clippy pass. Canonical Package run 34597972149 is active; duplicate run 34597976942 is cancelled.
• Deadmines source 393023c1, tree 93b2517aa7cb8c84902812e10c50ff26ac2e34d0, passed both final source reviews after correcting a test name and refusing conflicting repeated ADT placements. Root review SHA-256 1181b944d0fff5b51f720f7ca7b9f5ac3f30d3e0c1f6cd42399e93e1dd9d2b99. Independent review SHA-256 1da2fd45198d501a1b490a8625a1fb18c83e6230f284f40c10a49b89a88e31ea. These reviews do not claim native, archive or movement acceptance.
• Core #513 merged as 41e8af40b7646a91dc96ff5194c03108bbf81ade at 12:49 UTC after all seven checks passed. Final head 6c96c14c changes only test names after the accepted runtime input at 86e660f4; runtime equivalence receipt SHA-256 5288d3ee9d461dfef382c547832317e391d90d63d18fdad0fc9159f1c5cfa68a. Collection and release acceptance remain open.
• Core #515 merged as 375b38c3004b61500ce480a72a681d7999a5c9ca, tested tree 22fbc76a174a6f60abe2bea3d8148c047afff149. Native, actual Deadmines and outdoor evidence passed independent review a8d9ef9b0fa36b4c19e909c9560509016e2c8e5eb449cc69a24efeac0c39f8b4. The retained earlier center-floor failure remains diagnostic. Ordinary companion movement still needs execution.
• All four focused retained-target cases passed against Collection 87a47f9eee3bd5a3e3353f1ecea8acd827e0449c and actual Wasm 4b73c48d91008e7b0177cb37f3a7d8c887704aa41e523652868b23fec560bcaf. Core 91a6125a passed known-target retention and control-unknown cancellation. Caller-only Core cb2a7ea7 passed travel through ReadLimit with earned Quest credit and the complete live-target-to-loot sequence under the permanent crowd. RecordOnly staging proves no prior death, action, item, receipt or turn-in. The latter two cases passed in 155.47 and 9.80 seconds; the first includes an internal Wasm rebuild. All source, artifact, process and Package restoration checks passed. Independent Standards and Spec review 21de0bfd3b86cb6f8b13f19d9373fccc85561e0c4fc529e23cf0de38f3ab4e5d accepts these focused results. Full imported and attended acceptance remains open.
• The complete private Deadmines route passed at Core ccbe3ffb and Collection 590c2b1f in 41.77 seconds. All seven execution and capture rows passed; private cleanup completed. The final Package changes affect only Quest selection, leaving this companion path unchanged. Result SHA-256 a432681423bb0775872adb6ac717eb4431f3d3988a3c999448ac694ad2edec77; independent Standards and Spec review 83ff94b948cae2fa5d907690f58ea353a85dbd8642208367dc3006f4220981d1. This proves server-side movement and Transfer, not real-client rendering.
• The artifact build at Core 1e55a4c9 and Collection d8aa6f53 passed. Actual Wasm SHA-256 e0112c9fcba8d91153f1303bc19a7800736de3363cc7d3a31ee748d2b857a7bc. Its imported observation failed after 1,257 recorded passes without further earned progress. Independent failure review fe00986e57158dffed884f4428caafbd072688c445c3b965ccd446cecc446079. The original partial capture remains retained; all owned processes and private state were removed.
• Exact offline replay reproduces the imported route failures at the 4,096-expansion limit. Removing stale queue work alone, changing tie order and weighting the heuristic do not resolve them at that limit. Stale-entry skipping with a 16,384-expansion cap lets all six repeated 7-yard routes reach their destinations. A 32,768 cap adds no useful progress. Package movement reservations already use the Core cap. Calibration report SHA-256 5de876093eb694a7ebe6b2975ac44c183daec71d121669b267b1a643821695f3. The small navigation correction is being implemented; final merged geometry and capacity remain unproved.
• Both complete Collection d8aa6f53 runs closed with failures. Each passed 1,652 Module tests, all 30 class journeys, the ordinary Warrior level gap and all 27 Transfer cases. Composition passed 44 of 46 and durable execution passed 150 of 155. The retained captures identify scheduling, blank-query and wrong-observation assertions; scoped caller corrections have source review and await runtime validation. Full CI report SHA-256 d4976d79a87ea0cb20570330c609c707d394c84c9e194aca9bf768eaf086d910. The separate new Quest fallback regression stopped at an earlier boundary in its first attempt and is not yet accepted as a reproduction.
• The focused predecessor reproduces the exhausted-target loop at the exact Objective Deferred assertion. The corrected run then exposed premature deferral cleanup and a stale completed RecordOnly Objective. Both production corrections passed independent review in Core 0fdfc384 and Collection 827d02f6. The fallback caller now permits ordinary travel to the next giver and still requires earned Quest credit. Final source review SHA-256 bcd37fdbcf4648a2f34fa3ec84b737b4e6fc648d7415e661840b2f87ca8cb32b. These are reviewed corrections, not passing runtime results. The initial combined branch omitted the reviewed navigation production commit. Its build was stopped before any behavior test or Wasm build. Core 2256cc9c and Collection e5924770 now include the complete correction; both navigation file contents match the reviewed candidate. The interrupted build remains recorded and supplies no passing result.
• The clean Module and shared-code build produced Wasm e7899e610e3297a572bc927aa983f5333d1eab4491a76f0290e2e406e149128d, 9,670,125 bytes. The fallback case passed in 37.544 seconds and required ordinary travel, acceptance, turn-in, reward, receipt and XP. Seven focused checks passed; five failed. Result SHA-256 b480665478d3279581147e897ce6631a0f127687e6c35a76a14e235341a40b77. Source and private process cleanup passed. The final CI heads differ only in a test Boolean expression and workflow references; this artifact keeps its actual source identities.
• The five remaining focused caller checks passed after correcting fixture placement and observation boundaries. Final class acceptance then exposed an observer that waited for a Quest attempt after ordinary Objective replacement had pruned it. Collection 915aec37 recognizes the exact owning giver deferral after the second approach and restores the declared geometry fault. All 30 corrected journeys passed. The previous failed records remain retained.
• Final canonical Collection run 34667391293 passed all nine jobs at Core d183fe5b589983ddfb3909450c759a42c25f40dd and Collection 915aec37a2a41511aeabbe125d6f4fa2f694a407. Reviewed results are 30 of 30 unique class journeys, all 13 catalog Quests rewarded in each, 156 durable cases, 46 composition cases, 27 Transfer cases and one ordinary Warrior level-gap case. Every navigation-fault journey recorded two approaches, a 30-second destination deferral, alternate Quest work, geometry restoration and resumed Quest 783. Core Rust, Runtime Scripts and durable CI also passed. Independent final audit SHA-256 b08a3ca26d31c210f12a2aa524e9075b279e73f092f406155116ab40717613e8; extracted evidence manifest a17c503b77b236d73b6e34657835ffee189df4e3a430224d1acb076d4fd7c591.
• The final recorded build used the exact clean Core d183fe5b and Collection 915aec37 sources. Wasm SHA-256 255d628136241d18a1f607160e8f10627cb89350346d97f4e2e88d29f2b657a1, 9,670,108 bytes; build manifest 8e91a43fd3daf94d4edb2b6d0fa63aec3462865140c9deb8c92b50cdcdd512e6; build result b0e687fb439938c2e5337e809c827a06d32d470067d53ec896ac730ea1304808. All 39 enabled Package inventory files match the manifest. Build cleanup passed. The hour inputs are prepared; no imported runtime has started on this artifact.
• Independent review accepts RecordOnly fixed-fact comparison, freeze and resume, and both release suites with the new Runner selected. Criterion audit SHA-256 682f366b7f09d77a1a55bd68d1488f985edea39224616318511aef07407e58da. The prior 10, 25 and 100 bot measurements retain their original synthetic workload and source identities; they do not measure final imported gameplay capacity.
• The remaining review finding was valid. Collection 0f129f4de9aa9565daebe4c96d96c1c4f38cc094 adds the existing NAMED_LOOP_CONTENT constant to the named fixture identity check before any target mutation. Root Standards and Spec review and formatting checks pass. The correction is pushed for CI. Recorded 915aec37 artifacts retain their actual source identities and are not relabeled as this new source.
• Disk cleanup removed only unused npm execution caches. The active T3 process references match before and after cleanup. The new recorded artifact at exact Core d183fe5b and Collection 0f129f4 passed its build and cleanup checks. Wasm SHA-256 1daf043b22f050ac13ac39134d7859f426cc233e313ff93132ffe4414847db7d, 9,670,206 bytes; manifest cf06e8dd9cbe695bcd9402928333a4a6c2020f00c7e8f224788a91e98f231cc8. All 39 enabled Package files match. The unchanged reviewed hour driver started on private persistent storage. This records launch, not a completed acceptance result.
• The full imported run completed at 05:58:35 UTC on 12 September. It recorded 89,777 decisions, 61,414 completed movement legs, maximum scheduler lag 506,653 microseconds and no unprocessed bot. All 25 bots rewarded Quests 783, 7 and 5261; three rewarded Quest 33. Twenty-two ended at level 2, with one each at levels 3, 4 and 5. Capture archives finalized and private cleanup passed. These are observed gains, not acceptance of the remaining behavior.
• Independent hour assessment SHA-256 38f254fe4989e3021bde0fb6ee41e3749afa81be99bd91d5ad8b2965b382f648 rejects acceptance. It accounts for 136 no-effect gaps and 95 no-advancement gaps, including 21 still open at the final boundary. The trace records 13,747 QuestReadLimit and 1,184 QuestTargetMissing refusals. Geometry assessed 182,429 segments and found 80 blocked segments, including 15 ordinary Core splines. Behavior report SHA-256 1be628af5cf8c4c8786d4bf3258c689e04c02d2e25aeeab1cfeb8fc43bd6642b; geometry report 235f0993a5e337447f6946bea0103b4a22ed68ca505d45eb7ffcf4f56b4531a4. The existing checks and original failed evidence remain unchanged.
• The exact captured geometry replay reproduces the first blocked spline with the same endpoint and 5,930 route expansions. Core e10a8f0f089fc64956ec89219d8475a8f981634c validates the final post-collision movement prefix. Independent Standards and Spec review is clear, all 40 shared navigation tests pass, and all 15 original blocked splines retain clear shortened prefixes. Repeated planning reaches both observed goals without a shortening loop. Report SHA-256 0d1bbfb5e4de2a3effe790202a8b27262f83fcd7f1c33f5f24396f70e8df4646. This is focused correction evidence; final composition and imported acceptance remain pending.
• Two behavior callers passed independent Standards and Spec review at patch SHA-256 fd317c339e21cd3d97dc59cb2b1216742add32c73ffa5f13140ba8ffe349d1ef. They retain an actual secondary target during a permanent unrelated read limit and check an unchanged retry deadline, failure history and action count during repeated waiting passes. Execution against unchanged Package 0f129f4 is next. Source review does not claim a behavioral pass.
• Root accepts causal predecessor failures for both behavior defects. At Core b3c6382a and unchanged Collection 0f129f4, repeated waiting replaced failure timestamps and moved next eligibility. At Core c1944c58 and the same Collection, 96 unrelated rows caused an active alternative-target approach to become Hold with ReturnHome, while the Quest and Objective stayed unchanged and credit remained zero. The pursuit wrapper rejected an additional refusal-log marker because it captured the earlier Hold phase. Its final status remains 1; supplemental review accepts the actual causal assertion and snapshot. Earlier attempts failed during setup and supply no pursuit result. Root review SHA-256 b726e61a4ce1d53b2014c6ada279a4f749de7ddff91a1c129d8317e3b887a7ce. Corrected production execution remains pending.
• Focused GREEN v2 passed both exact cases at Core e30cac702ebcebd53e1bf19cafc1a93350a5ae16 and Collection 684d8910d3a7a638670d5d824e92a9faf6465b40. Active-target continuity and earned credit passed in 47.043 seconds; bounded waiting passed in 10.273 seconds. The first corrected attempt exposed Provisioning clearing active Recovery, which the final candidate fixes while excluding only Provisioning time from false stall accumulation. Independent review accepts the diagnosis, correction and final results. Result SHA-256 10f9163f54f943100bbc914b028c1adf2ea4df9b2a5c9e25c047fa4a38a123c5; Wasm d3d3033da90be85ba12317fb2ead2d4f40ee03d3a751a0c32d29a61802a0efcf; manifest e71f26642fa1273a6a63cc0373f38028c3115746d7935e72e30b706bc7844eb0. Sources, process cleanup and Package union checks pass.
• Final composition preserves the reviewed navigation and tested behavior changes. Its recorded build at Core 85c2c204ba0050c2b1ad31477f722233494d304d and Collection e9309134b995445a85d5260629065c7e1f80ae04 passed. Wasm SHA-256 7aa8ca217f90b1fad5131fbc7c352e5e8212555efc9f8aa22372d1463416fbfd, 9,667,742 bytes; manifest baf121a116d53ee79c4e2eb121b35598282b67cd116b4fdeaf38f5fffd08d287; result 4b076e25d5d5902dd429b0d8fe12867359ae6f39248295c406f929997f398884. All 39 enabled Package files match. Current heads differ only in equivalent navigation-test literals and workflow references. The next hour is prepared, with input SHA-256 aa67250e658aa61b6a8073f5c20e6dcbf8bf2ba5db732d3c9506b8f03687fbfb. It has not started.
• The unchanged imported-hour driver started at 08:39:58 UTC on 12 September with recorded Core 85c2c204 and Collection e9309134. It owns a private loopback runtime and persistent storage. World content, terrain and navigation imports passed; collision geometry is preparing. The 25-bot workload, warm-up, 3,600-second measurement, resource bounds and behavior/geometry verifiers are unchanged. Current remote CI runs independently. This records launch only, not imported acceptance.
• All imported preparation steps passed. The observer started at 08:51:51 UTC, followed by its unchanged 30-second warm-up and 3,600-second measurement. Capture files are growing. This remains an active run with pending behavior and geometry verdicts, not an accepted hour.
• An independent early review of complete saved pass records reached 783.574 measured seconds. Seven of 25 bots had rewarded Quest 7 and Quest 5261; one had rewarded Quest 33. The three bots with zero Quest 7 credit had recent combat and movement, with no established persistent pursuit loop. Repeated QuestReadLimit events were about 30 seconds apart. Review SHA-256 35c5bce5d9073eb5b2bd95bc0e2588ca732c0193a4b90aea162344cc7c23a44c. This supports continuing the observation; it is not a full-hour acceptance verdict.
• Canonical durable job 103523882025 started at 08:10:54 UTC and reached its declared 60-minute timeout at 09:11. Its full log also records playerbots_active_quest_overflow_preserves_the_retained_purpose failing at quest_catalog.rs:1207, before the timeout. Source diagnosis shows objective reconciliation copying a Quest refusal deadline into the shared eligibility time before arbitration, which delays an otherwise eligible Defense retry. The correction will assign the stable Quest wait only to the selected blocking Hold. The original failed caller remains unchanged. No complete durable-suite pass is claimed.
• The next independent saved-capture review reached 1,683.515 measured seconds. All 25 bots had rewarded Quests 783, 7 and 5261; four had rewarded Quest 33. Active Quest 33 bots had useful movement or combat in the preceding review window. Repeated QuestReadLimit records remained at least 30.1 seconds apart. Review SHA-256 1bd744ed05a72bd061461fb85ee20132981ae66ef7c40e00fcd9bc0c6343fe82. The observation continues; full-window behavior and geometry acceptance remain pending.
• The reviewed overflow correction is pushed at Core 0a7e123e3fe43162f2c602290e194c927566b9a2 and Collection fe8b01b3082d00f0df29afa20d75cb9fb6c769ae. Source patch SHA-256 9e49a1fd723c59ffa7700600b5c7ba12ebc9fc460bb3e9b1697188eb16eea849 and caller patch 9bcb09a71d69a0db2fd0ad62c14c778be942d0f25c4d9bd779ba75e5e24543d1 passed independent and root review. Two waiting passes must preserve the exact Defense retry owner, count and deadline, then the original due-Defense assertions must pass. CI runs 34685881745 and 34685879502 are active. Runtime validation of this correction remains pending.
• The corrected imported hour completed at 09:52:22 UTC on 12 September. It retained 3,600 measured seconds for 25 bots, 89,921 decisions and 38,580 completed movement legs. Maximum scheduler lag was 514,616 microseconds. Every bot was processed and showed authoritative progress. Both archives finalized without errors, and private node shutdown and directory removal passed. Run SHA-256 04813955217885176ea3c6656eece0d106b8ef27b4d8b41fa602770ca664b544. The artifact retains its actual Core 85c2c204 and Collection e9309134 identities.
• The unchanged behavior verifier found zero refusal-spam candidates and 505 intervals needing causal assessment: 406 without movement, damage, XP or Quest progress, and 99 without XP or Quest advancement. The longest advancement gap was 2,860.161 seconds. Behavior report SHA-256 3ce0d4986119e2594b524f50a9c7b5db77da5076cf2e9d3e358b65f64ce87c77. Geometry assessed 112,823 segments and flagged six consecutive entity-position segments; all emitted splines were clear. Geometry report SHA-256 61b7678d5e9dc707e3ce14fda3704fe058807e0f3bc7fea0a39d9076e49d5978. The original failed findings remain preserved. Independent behavior and spline-lifecycle diagnosis are in progress; no imported acceptance is claimed.
• The recorded build at actual Core 0a7e123e and Collection fe8b01b3 passed. Wasm SHA-256 74d879204115d460a93447b088b02fc8e91d7486f4107ae323435dfa9b5cb83d, 9,892,834 bytes; manifest 81afecccf9dea7f85615e3add2418478e94db4e00c9dbdb74c0d415299021ea1; result 6c3525a514131006092c677f1b8d0a1ec100b59a136ea60ad879d1f358c72c03. Stable waiting passed in 10.526 seconds. The overflow check failed before its new retry assertions because its initial Cohort pass had already started combat. Root and independent review clear caller-only setup patch 33290cc04b4e8a863b044adcf246fa943902a89471703c2cda6bc5fa40273c5d. It discovers the retained Quest under RecordOnly, then selects and parks Cohort before overflow staging. All original assertions remain. The failed case will run against the same recorded Wasm.
• The corrected overflow caller passed 1 of 1 tests in 119.339 seconds at Core 281bc0e22006a0ec1a12a13a7a80e7e1765e1ef6 and unchanged Collection fe8b01b3. It used the exact recorded Core 0a7e123e and Collection fe8b01b3 Wasm, SHA-256 74d879204115d460a93447b088b02fc8e91d7486f4107ae323435dfa9b5cb83d. Root checked the caller-only source scope and actual test result. Result SHA-256 5f2b092aa6e2b2b5456e33d8c1efa02dce72b60c61f9179d6bfc60803f56033d; test log 7c191aaee42a9036c916403bfbf000dc8a31c7cd70f2bab302e0d530ca70449f. Source, Package inventory and private process cleanup passed. The earlier failed setup is retained.
• Independent replay of all processed snapshots in the 14 unfinished advancement gaps found 3,983 Quest 33 creature-target snapshots. Of these, 3,962 shared their target with another ungrouped bot. Each stalled bot selected 4 to 13 distinct targets and moved across a 94.1 to 186.0 yard coordinate envelope. The longest gap lasted 2,860.161 seconds. Source prefers the one catalog target before an active alternative and treats foreign-tagged creatures as valid autonomous work. The trace does not include private Loot Tag rows and cannot name each owner, but ungrouped attackers cannot all earn the same tag-owned reward. Root accepts this as a productive-target selection defect. Behavior acceptance remains failed. The correction must respect Loot Tag ownership for autonomous Quest and Grind work, preserve a useful active target, and leave Defense and human-directed assistance available. Final independent assessment SHA-256 7cc9da603b061132121ad2784c2b4e777e40b6333392ae6864439de2bdd6dc72.
• Independent geometry diagnosis reproduces all six flagged updates as exact Core f32 interpolation on clear retained splines. Both endpoints are walkable. The reviewed checker preserves ordinary segment checks and resolves a blocked rounded chord only with exact parent identity, simulation timestamps, reproduced endpoints and clear parent geometry. Its 16 focused tests pass, including refusal of off-path, reversed, missing-clock and blocked-parent cases. Verifier SHA-256 660c2cff67da372b0c9c560cc5ce672fc2139dbb3189f8846d1563100b33a85d. The full saved-hour assessment is clear: zero blocked paths, missing segments or unexplained discontinuities, with all six original findings retained alongside their exact proof. Result SHA-256 d7562c7de2f0ff5b5da065fb68ec96682cf0c95dd82abb31c73c48d67d736e80; diagnosis 72c40f3e7f265a747aaedcd8542c0e6977ee462a5071daa033142bf0f3e5588d. Root verified all six original identities and blocked cells remain recorded. The original v2 failed report remains unchanged. This accepts static geometry evidence only; behavior still fails and visible client motion remains pending.
• The Transfer caller correction passed its one exact case in 116.84 seconds at Core f72475f0a57d22ca4cb10a760cb0b27fbaa4bd24 and unchanged Collection fe8b01b3, using the recorded Core 0a7e123e and Collection fe8b01b3 Wasm. It waits for the actual AreaTrigger approach before the original unique-action, CannotAct, objective and empty-intent assertions. Root and independent source review are clear. Result SHA-256 19de8aeb1bed4631c872d3842f9c772f805ef9e23a82b1cdc560a30826269bdb; log 2d3a66d0ec0404cde3356241b6aaa8530090b3c158f1a600bc439bfde6dfcb43. Source, Package inventory and private process cleanup passed.
• Canonical Collection run 34685881745 completed successfully at fe8b01b3082d00f0df29afa20d75cb9fb6c769ae. Root verified all nine jobs passed. Core CI also passes at 0a7e123e. These are the current remote sources before the productive-target correction; they do not resolve the failed imported behavior result. The final correction will receive its own complete CI and evidence review.
• A focused Quest case reproduces the productive-target defect at Core 45f6469156185d84e160eb7db0023ba0552df6c3 and Collection d16f7e0192205004ef8375b47653e80da95984ad. Production behavior is unchanged from the current PR sources. A second ungrouped Character creates the live Loot Tag through real positive threat. The subject selects that foreign-tagged target while an eligible alternative exists. Root accepts the saved state as causal failing evidence. Snapshot SHA-256 4ce20028c864a6271ca1ca36cd93c34abfdda9047423751b9ea8eded7af16e96; log 13084f0be7ef69a5032203425d7c4c7044a6110f82eedb7c121d1023196e2c43. The paired Grind case failed setup because normal spawn scatter invalidated an exact position assumption. Its original failure remains recorded. A reviewed caller correction derives target positions from the actual frozen subject and records the exact new target identities; only Grind is being repeated with the same Wasm.
• The productive-target correction passes all three focused cases at Core 8692be3d15cb0eda70ddd9d1cbd73486f7f3ea8c and Collection c46e9e3ec62434cca60eb8d69984d11bc918542e. Quest passes in 50.038 seconds: it avoids a foreign solo tag, retains an eligible alternative through primary respawn, switches when that alternative becomes foreign, and earns ordinary Quest 7 credit with sole corpse entitlement. Active Grind passes in 9.016 seconds and cancels its old foreground approach before switching. The Core Loot Tag fixture passes in 4.266 seconds, including party departure and revocation. Root and independent Standards and Spec review are clear. Result SHA-256 181f0b5d4059507c5799eb43402979c436f04fde1f88ca3cce893f4adca6c916; snapshot receipt f619b1c04ed9284fd840386b66530fc4d2c4b3ba138c330f857c9dbb8fdca431. Build, exact Wasm, source, inventory and private process cleanup passed. Both earlier causal failures remain preserved.
• A fresh imported-world run has started on the recorded corrected Wasm, SHA-256 3f49786c9c86591e8cef75a7d6e06dd13074b39b3b3eb45744ce5cb34564389e, 9,662,921 bytes. Build manifest SHA-256 256f4550a177f424693cd2d9228a3a4e38f7f62af2a7e93e4563559cf8c8b6ec. The 39 immutable Package input files match that manifest exactly. Reviewed run inputs SHA-256 3c9993dfd2198580edd9c779de108525c82f7a6ff944715163ee824c1336a563. The imported content, geometry, starting positions, 25-bot roster, 3,600-second measurement and behavior criteria are unchanged. The accepted geometry v3 checker remains in use. This is a run in progress, with no new imported acceptance result yet.
• Core 8692be3d passes all current PR checks, including Rust, Runtime Scripts, automated review and durable Module and Gateway tests. Independent durable audit confirms 94 Module cases, one auction Refusal case and four Gateway ownership and projection cases. Log SHA-256 9e885284ae95d73daa4aea570345977c6aa7d174c015427bef9a905b37482698. Package Mage, Priest, Warrior, Transfer and composition checks pass. The full durable suite failed its new Quest caller while waiting for exactly one credit. Retained state had ten credits, the Quest rewarded and later ordinary progression. The exact earlier selection assertions passed. Independent review found that changing the polling predicate alone cannot preserve the primary target Action and corpse entitlement after a delayed pass; that draft is held. The correction must retain those postconditions. General Package module and datascripts jobs depend on the unmerged Core Loot Tag API and will be repeated after that merge. Core remains held for imported gameplay acceptance.
• Shared tick, movement, survival and Quest helper extraction is prepared on isolated Collection commit 215cdc9aee9d6f1d761be9154182f6e2b601a7e7, based on ddb01ed404175f1cc111de8768bd8d6567363922. All 15 moved function bodies, six constants and 16 tests retain their original contents. Root and independent static Standards and Spec review are clear. Cumulative diff SHA-256 bf7735071855ba6b7caefe1f94ac2b43da9325b9c717c97564d7a8ae4e65e6ca. Both controllers and their registrations remain. The branch is unpushed and has no build or runtime acceptance. The Core comparison test still requires an explicit source-identity adaptation during retirement. No observed artifact changed.
• The corrected imported hour at Core 8692be3d and Collection c46e9e3e completed all 3,600 measured seconds with 25 bots at 12:16:47 UTC on 12 September. Owned-node shutdown and private-directory removal passed. It processed 90,444 decisions over 7,181 scheduler passes with maximum scheduling lag 508,838 microseconds, zero lost due bots and zero reducer rollbacks. Peak Standalone RSS was 878,276,608 bytes; minimum free storage was 9,549,811,712 bytes. Compared with the prior full hour, bots at level 3 or higher increased from 9 to 20, Quest 33 rewards increased from 8 to 14, and unfinished advancement intervals fell from 14 to six. This run still fails acceptance. Behavior analysis identifies three RecoveryCapacity Refusal bursts and six unfinished advancement intervals. Static movement has zero blocked or incomplete segments, but two ghost relocations are real defects: a retained zero-duration stop spline returns the released ghost from its graveyard to the death position on the next tick. The original geometry result remains incomplete. Run SHA-256 2996e03e22bd47df40bfb3f7438b074b88888f9648ceb0cfa8ca4a69c4ec7e11; behavior d6635be9d88cb5a5d79da548f8761062c692008efc0c4e72a1a184668086b80b; geometry c195f098f886110d40517c6b46c151eb603c1a60453697f9756384747f2888da.
• Both imported behavior defects now have reviewed corrections. Core teleport will replace any retained movement leg with a stop at the destination. Package recovery will record RecoveryCapacity once while preserving one-second reevaluation, newly available healing and ordinary expiry. Relocation regression is running against unchanged production source. Mandatory Cargo builds changed Wasm identity from the recorded hour, so earlier attempts stopped at the identity check. A separately identified RED baseline now records its own binary, toolchain, paths and exact production-source equality. It is not the observed hour artifact. No GREEN runtime result is claimed yet.
• The next hour observer has passed root and independent source review at a0c789e0cb99e5ea43eb92b5c5333b16ecdffb18. It retains five tag, party and corpse tables plus creature and spawn rows for the four compiled autonomous target entries. Declared NPCs stay outside bot movement and progression counts. The 13 queries and their schema are fixed; existing gameplay criteria and archive limits remain unchanged. Offline checks pass: the owner-byte Gate, three observer-evidence tests, four behavior tests and 17 geometry tests. Rust build and runtime remain pending. Versioned driver v7 SHA-256 446fbf0e0edc4b0426f8c8096b506d379aa69a15833239eb7cdb267100418dd3; behavior v4 f6a6d865dcbfa589e88d4c476bcbbb735647ddd0a4c572d4f003ca1599cd7a4f; geometry v4 ca714e1e8a1de1638147ff706bdfeaab40403cca6ad2ed42a87350ada5e010c3.
• All three relocation regressions reproduced the old defect at the retained movement boundary on the separately recorded RED Wasm bfb2af33041423eea93aad2bfea0ddd276715b0761e96b136a3737e0429e6cf3. Same-map teleport retained its active old destination; graveyard release retained its old death-position stop; cross-map departure retained a source-map leg. The wrapper misclassified multiline libtest failures, so its original failed result remains unchanged. Root verified the logs and snapshots separately. Assessment SHA-256 227f7027f21d469a172c20223c6445068ea154771eb4f87af3d9225702d87d1c. Reviewed Core correction 06ee51be4e2d57ee17be5e336a5ad3cee8e120b3 is committed. Its movement-tick, resurrection and cleanup checks still require GREEN execution.
• Recurring temporary-storage quota failures stopped recovery test preparation before gameplay. Five finalized older evidence directories, containing 489,963,101 bytes, moved to persistent storage after content and mode verification. Their original paths remain usable. The active build reported 313,075,712 bytes of quota headroom afterward. Current source and PB012 artifacts were untouched. The next compile found a missing table-trait import in the new recovery fixture; its reviewed correction is Collection 141341b37abcc4b44cde41c5431accfd820320c7. Recovery RED has restarted with the same behavior assertions.
• The recovery regression now reproduces the repeated Refusal on the old production behavior. The first wait records one RecoveryCapacity failure and retry count one; the next wait records two of each. Both keep exact one-second reevaluation. Root verified the snapshots, recorded Wasm and cleanup. Result SHA-256 0339e3eeb9d340e17a8e49acb6f767203ce7f6eb4597201aa1578ec920845682. Earlier fixture setup and compile failures remain separate from this causal result.
• Combined Core 607ca975a0b6aef95998a3569cae09454e205df3 and Collection be599eb05e1d2bfa84773550f2af2af0d334c244 pass independent Standards and Spec review with zero findings. The build includes destination-bound teleport stops with distinct spline IDs, one retained RecoveryCapacity Refusal, completed self-heal evidence before ordinary recovery expiry, and the exact PendingTransfer party-mirror fixture correction. Review SHA-256 146b27e4ca64da7e03f2b0b60f6879b825609b9f00cb144d59b16ed454673eea. Six focused runtime checks have started; no passing result is claimed yet.
• The new ownership interval assessor passes seven focused checks and independent source review. It preserves exact selected actions and candidates, captured tag rosters, current party membership, target life state and corpse entitlement. It distinguishes facts after a transaction from decision-time facts, and it does not accept an interval merely because progress follows. Assessor v2 SHA-256 9b8035d87ba459b44ab6a305b5e46505c417edaf422d0ef548e74a5270707376. Captured windows and boundary anchors do not prove state continuity through omitted gaps.
• Combined validation at Core 607ca975 and Collection be599eb0 passes all three relocation cases and both Transfer cases on recorded Wasm 2281bde157d055226057f1370f978c374faeb589d77992fadcb672659219a6dc. Same-map relocation and graveyard release preserve the destination through the movement tick and spline removal; resurrection stays at the graveyard. Cross-map departure and Gateway party-mirror recovery pass. The recovery test passes repeated-wait assertions, then fails because its Priest has zero mana for a 30-mana Lesser Heal. An existing fixture setup supplies mana before that step; all behavior assertions remain unchanged. Only recovery is repeating. Original failed result SHA-256 76ee1fcde80ff5826f7a2941bfd83b105280774ad76650db3e9eac23816ef915; independent partial assessment 6c2275cf85c9f296898415c8a2b090b54081dc95889924ee5dfb31763bb0fe49. All case logs, capture hashes, Wasm identity and private cleanup passed review.
• Recovery now passes its complete focused case at caller Core 6933cbb5bdd20379feac4ab301840cb9bf0ca01b and Collection be599eb05e1d2bfa84773550f2af2af0d334c244 on unchanged recorded Wasm 2281bde157d055226057f1370f978c374faeb589d77992fadcb672659219a6dc. One Refusal and retry count one survive repeated waits. Self Heal 2050 completes with matching cast ID, CastResolved Action and cast progress; no pending cast remains. Recovery stays intact through completion. After the ordinary decision deadline and oldest-attempt expiry, the bot selects Move.Home with unchanged failure history. Result SHA-256 c291c45f2e28d754df690fe82387d8a7d0200a52463650e00e37d947dcbb2635; final snapshot b12a084557b320f1a0314e78708c7887eb6412d20622517616843d64fad2a2fa. Root accepts this plus the five passing combined checks. Independent assessment 98edbdca374f0f17286b97dfc7a87383bc699db2072e9f98be5411ac97c0f7bf. The earlier failed callers remain recorded. Production source and actual build identities remain Core 607ca975 and Collection be599eb0; the later Core changes affect only this test caller.
• The target-ownership observer passes local format, Clippy, all 34 tests and release build. Its Ubuntu, macOS and coverage CI pass. Observer PR 18 merged as 4fa71ac357de847c9be1b665015d4e62df3f31e5 with full tree equality to reviewed source 27d086512fe767e321faff1a7bac99f02a5f6fc3. Retained binary SHA-256 b0b78df79cf43b4c79d1363b2f5cc1598d1490624a689195aba99072cf92e92e, 1,628,176 bytes. Native result e8931c3ad2ae63461f9eee24323ff95541f27c94004785a643ac5d3e292c45c9. Root verified each command and output log, source identity, binary and cleanup.
• The next imported-world run started at 14:16:27 UTC on 12 September and is preparing the same imported content and geometry before measurement. Its reviewed inputs bind recorded Core 607ca975, Collection be599eb0 and Wasm 2281bde1, 9,668,912 bytes, with the new observer and versioned ownership-aware analysis. All 39 Package files match content identity 3040d572f55a283e873071cb3c6e43be032b87e2aad793ff9a112b4e2557009d. Manifest SHA-256 e24d5e28a269b9e0ceac49c26b593c9da14ba5e4d366886a9a5fffab0253c5b2; independent binding b6d7a5661cb60c02c0983c8cd623fc29ef7d714278f6d4ef322f154e2bf2d695. The workload remains 25 level-one bots and 3,600 measured seconds with automatic provisioning and ordinary XP and Quest credit. No new hour result is claimed yet.
• Final review follow-ups are pushed at Core 2e078e60db80213e3654895a14ebbde4a85b3e05 and Collection 1c0f7afde2ad5b1aea2a7d1cf2578fe429c89f94. Core changes since the recorded build affect three test callers only. Collection changes affect its workflow only, including a separate pinned policy-comparison checkout that leaves predecessor build inputs unchanged. The playerbots and dungeons trees exactly match the recorded build. The policy artifact now states its health normalization, and a route deferral assertion starts at the proven revisit. Independent Standards and Spec review is clear. Final CI runs without changing the active imported-world artifact; general Package checks still require the Core API merge. The attended handoff records final PR source and actual Wasm source separately.
• The imported measurement began at 14:28:27 UTC on 12 September and remains active, with closure scheduled around 15:28 UTC. Recorded memory and disk use remain within the unchanged limits. Final Core 2e078e60 passes Rust, Runtime Scripts and durable CI. Package composition completed 45 of 46 cases; the last stopped on a missing historical comparison checkout. The reviewed workflow correction is pushed as Collection c4960b5be7b187cb016084c65ebd92cac59ee14a. Production trees still match the measured build. New canonical run 34700753123 is pending; the failed run remains recorded.
• The isolated Package retirement preview 848d39c8 and Core caller preview 0b8b04e0 have clear independent behavior and standards reviews. They remove the old executor, bot-led grouping, personality artifact and obsolete callers while retaining serialized compatibility and bounded migration. Final composition must carry the current comparison workflow, rename the moved target-window constant and include the separate Architecture Test. These are source previews only. They do not establish runtime, attended-client or Realm cutover acceptance.
• The full 25-bot, 3,600-second measurement and final geometry collection completed on 12 September. The capture records 7,175 scheduler passes and 90,294 decisions. Peak Standalone RSS was 869,535,744 bytes and minimum free disk was 7,065,395,200 bytes. The driver then exited 1 because copied read-only Package directories prevented private fixture deletion. Root verified the owned server had stopped, restored owner write permission only on twelve private directories and removed the fixture. Original run record f63345b24cbf7f6819ab45357b43e958b6288316fa06fd6892612122552a0dbf and traceback 44e142e48c85686eafb8086eeb988dfe874d1e6131932b958c3dafc73f8429f9 remain unchanged. Cleanup receipt 69143e41f4b13066f78eb988d5976fdeafc14cbfe6368495b22ab9508827b775 and separate finalization d178c10b048f441fe912ea7f88cced93d5b1fb76937ffe9ec7e2364f71589ebf bind the repair. Independent review accepts this post-measurement cleanup evidence. It does not establish gameplay acceptance. Versioned analysis preserves the original capture and thresholds.
• Current Collection c4960b5b companion and lifecycle composition CI passes after adding the missing comparison checkout. Transfer also passes. Remaining current-head durable and class jobs continue; general Package checks still depend on the Core API merge. The paired retirement previews are now Core a4aedc2e7498b1e9e4b81719599f5bb25c0c7e09 and Collection 94417428d9a84451db65ef7501839f8353ab05d7. Independent source reviews are clear. The Architecture Test rejects retired executor paths, active goal access, bot-led Group Intents and removed personality scripts while allowing exact compatibility owners and excluding test-only code. Compiled validation is underway. No retirement delivery or Realm operation has occurred.
• All three analyses of the completed Core 607ca975 and Collection be599eb0 hour are closed. Static geometry is clear across 110,982 assessed segments, with zero blocked or incomplete segments and zero unexplained discontinuities. Geometry SHA-256 7114e33e1d2d0f7845eceddfda69576834179991214b7211657105a2658cd0e3. Behavior records zero Refusal-spam candidates and 441 intervals: 327 without movement, damage, XP or Quest progress, 111 without XP or Quest advancement, and three injured-healer intervals. Behavior SHA-256 6d25e73c0e132591606f3314d2c1136f9fc09cf634c87f1004e9881253b2e584. Ownership SHA-256 d44db9d4080ba9c102c249486387eee535f3c85f3ca15d6cd22ba4d010cb9192 retains all 441, seven open at capture end, and 35,320 state episodes. Independent causal review is pending. Final levels are ten at 2, four at 3, four at 4, five at 5 and two at 6. These counts do not establish acceptance.
• The paired retirement Architecture Test passes all four cases against Core a4aedc2e and Collection 94417428. Changed-Core formatting and all cleanup checks pass. Result SHA-256 768edec95c606eb55778a5b66cec8c3035e7cf267cf6cb98b1650454d5d8c847. Focused strict Clippy fails with six removal-related diagnostics and seventeen retained findings. Root is removing unused Package code, limiting retained Group Intent fixture writers to debug reducers, and preserving the documented shared Script Answer API. This is held source preparation, not retirement acceptance or delivery.
• Independent causal assessment of the full Core 607ca975 and Collection be599eb0 hour is FAILED for imported behavior. All 441 intervals are accounted for: 420 closed intervals contain synchronized Quest fallback selection, seven open Quest 33 advancement gaps contain the same pattern, six are bounded waits, four finite pursuits, three resolved healer intervals, and one Grind interval lacks entry 721 target facts. The source always selects the nearest Quest fallback, while Grind already varies its choice by Character among three nearby targets. The proposed correction reuses that policy for Quest fallback and preserves active Fight and catalog-source priority. Assessment v2 SHA-256 47a8a266458a86944ee09c0e1e8e30899d5116e98ba4c4ae064aa40dff8b6107. Geometry and Refusal stability are clear. The seven open gaps and reward counts describe the result; they do not introduce a new all-bots Quest 33 requirement.
• Current Collection c4960b5b canonical run 34700753123 and push 34700750711 finish with all seven gameplay and firewall jobs passing: composition, durable, Transfer, Warrior, Priest, Mage and firewall. Only general module and datascripts checks fail at the known unmerged Core API dependency. They must pass after the Core merge. The next gameplay correction will receive its own checks. Held retirement Core 7304427d and Collection 1218eb83 pass Architecture Tests 4/4 and debug-reducer compilation. All six removal-related lint diagnostics are gone; seventeen retained strict-Clippy findings remain. Native result SHA-256 96e4eac31e63d35c197141efa8c22610d1b03a8645e0a8295b5fcaca85509663. No retirement delivery or Realm operation has occurred.
• The causal 25-bot regression fails on unsalted Quest fallback with all subjects selecting one alternative. The one-call correction passes with choices distributed 8, 10 and 7 across three eligible targets. The unchanged productive-target case also passes, including retention through primary respawn, replacement after foreign retagging and ordinary Quest 7 credit. Recorded Core 24ab2c71 and Collection 2fdc43d0 produce Wasm SHA-256 07b74b12851aa15f05d75905d6b66f7cbbafe0bc5b977630b628154c10488ab7, 9,665,319 bytes. Result b446d6fc2687181cedfe3f5ca098becc1bbeb1e98ac046ffda1458d96b40c746; independent Standards and Spec review 578140feaa3e6f5acefe812ae52585cfd8a2fed447c38ab7cdaaa63a1c433a15. Final Collection 6fb26891 changes only four Core CI references. Both final PR heads are pushed and CI is running. A fresh imported hour remains required.
• Observer PR #19 merged as f9c23ca8c6223a74032cc04a023948617e103c1a after formatting, 34 tests, Clippy, release build and all remote checks passed. Its merge tree equals reviewed source ac10282d. Exhaustive review finds entry 721 is the only missing autonomous combat or Grind target entry among all 441 prior intervals. The raw selected field also includes Quest givers and self-heals. Nearby eligible Grind creatures 257, 883 and 890 require capture coverage before the next hour. No creature eligibility or acceptance threshold changes.
• The source audit covers idle spawns, six-yard random movement and waypoint segments against all 25 home leashes. The complete autonomous Grind entry set is 6, 38, 69, 257, 299, 721, 883 and 890. Observer #20 captures their health, life sequence and ownership through 21 fixed queries. Source review, 34 native tests, Clippy, release build and all remote checks pass. Merge 23913d118fc7e653886be1c2ed6c87e623a4ef18 has exactly the tested source tree. Binary SHA-256 c42f315861d93fdbc061a39cbe6cd5558ecc89cc5499cea4ef6746a303b7474d. Scope review c74014b5733832e2423b1b7f9b7cdb3ab08fe67cfca7d3db10218b545e9a3933; tool review 0efaa3ccb750c2b72b6ee9e37deb8795c1b90c320aa8d44314a7434bc3ccb24c. Behavior, geometry and archive limits remain unchanged.
• The fresh private imported run started at 16:53:51 UTC on 12 September. Reviewed inputs SHA-256 ded35f398f18497086e438e8dbb4034378f2a4d891b49b2beb19fae6196ae3d2 bind the accepted Core 24ab2c71 and Collection 2fdc43d0 build, its 39 immutable Package files and the new observer. Independent input review c1629c73caf95e50de432b7917c22c0e7187953e00af6e62318eacf52f6f0fa1 passes. The World Shard import is in progress; the 3,600-second measurement and independent behavior assessment remain pending.
• The approved dispersion correction also composes onto the held retirement previews as Core ce2e096b and Collection bd636f05. The caller and fixture patches match their accepted originals. Only the moved local pick_salt path and four CI refs adapt to retirement. Independent Standards and Spec review 0acc9223676e0bce771dd3161dbf8c7ca3add7646701ac22d33927a906971a75 is clear. This is source composition only; its native checks, retirement delivery and attended acceptance remain pending.
• Import, navigation, collision and staged-roster checks passed. The observer started at 17:05:43 UTC and the full 3,600-second measurement is running with 25 bots. A live capture prefix shows all 25 opening Quest rewards and initial Quest 7 kill credit. This is progress information only; no final behavior or capacity result is available. Core 24ab2c71 now passes all current checks, including durable run 34705922518, Rust run 34705922519 and scripts run 34705922524.
• The fresh imported hour completed at 18:05:44 UTC on 12 September; geometry capture and private cleanup closed successfully at 18:06:27. It used the recorded Core 24ab2c71 and Collection 2fdc43d0 Wasm with the complete eight-entry observer. All 25 bots reached level 2 or higher; 16 reached level 3 or higher. The behavior verifier records 558 intervals for assessment and zero Refusal-spam candidates. Eight progress intervals remain open at the end. Imported behavior therefore fails acceptance; later progress does not explain the preceding gaps. Behavior report SHA-256 d77b09a6873fcc4ccf562b4c9b8d27404d1140457dc1a895d0c6f916f92cdf8f; ownership report 72948545820c69c3bbf4352695074b8b91f0dd4f632e974b5c2574705291fe94.
• The completed geometry assessment is clear across 108,175 segments and 113 terrain cells, with zero blocked segments, missing geometry or unexplained discontinuities. It accounts for 38 Release Spirit relocations. Geometry report SHA-256 12dbeecd8508520845af5c9e05465daaabb49b7d40b990b070722a2cb2b1b9bb. The four injured-healer intervals contain active work and five resolved self-heals, with no idle selected action. Focused healer assessment 4d265f7af17af6cbac08e0917bdb03dab90a58be638a62c696014a1117a854ca. These findings do not establish rendered-client acceptance.
• The first offline behavior invocation stopped before archive analysis because it expected the preceding failed driver cleanup receipts. The corrected verifier accepts the actual successful driver cleanup and rejects partial, false or mixed evidence. Nine behavior and eight ownership checks pass; independent Standards and Spec review is clear. Raw input, original run and capture bytes remain unchanged. The hour was not repeated. Analysis transition receipt SHA-256 2479be41cd80b0997b073e64725ca7f45048ea5e40ed31ffbed84bb9b4a2cb34; review ed7a9cf4c4a62dbda424ad060700380d4f3cc9ef3457848b376c6cd01f4bbaf8.
• Source review finds that an inconclusive bounded corpse search stops CreatureLoot before a separately valid live alternative can be considered. The proposed correction accepts only a fully validated live target and retains ReadLimit for every other outcome. Search bounds, corpse entitlement, active Fight retention and combat checks stay intact. Source assessment SHA-256 27c53375e5805766afdbd843b1a591ede394d8ddef8485cb29896e2e4e6dbc4e. A focused real-gameplay regression is running in isolated worktrees. The complete mapping of the hour to its causes remains under review.
• Core #516 merged normally as ba6f1ba57b672c7c468a6a9786d02f4fdef84cdb. Core #518 then merged as fd2687f3056f2964d75bbd483bde68a463499434. Every required check passed, and both merge trees equal their reviewed heads. Their additional changes over the recorded Core build affect only the recovery test caller. The current Collection a88e828a changes only four CI references from the recorded Package build. At 18:22 UTC, its module, datascripts and firewall checks pass in both current runs; gameplay jobs continue. Collection delivery remains held.
• The held retirement composition at Core ce2e096b and Collection bd636f05 passes all four Architecture Tests and debug-reducer library compilation. Strict Clippy retains exactly the accepted seventeen findings, with no new diagnostics. All source, Package union, process and temporary-directory cleanup checks pass. The orchestrator independently verified all fifty evidence files and the test executable. Result SHA-256 9a31ff37fdff0f31b528f4ac650088c804995f86a75c0d1afd1e51526c1ce3b8; evidence manifest 447763715f8f94bb438596e8be568f679fa990fb30f248877b58270c37c6c1d4. Retirement delivery and activation remain pending acceptance.
• The collection-search regression fails on the old production path at the intended combat-start assertion, then passes with the bounded live-target fallback. Real spell 133 kills the alternative; completed OpenLoot and TakeLoot produce an exact eight-item receipt and one Quest 33 turn-in. The unchanged ninth-inaccessible-corpse case also passes and retains ReadLimit without an invented target. Core caller 2819939d and Collection 97f29a23 are clean. Focused report SHA-256 6849cecb6eac1e2c25368708d4d153726270deeeb9e8e871a719dfd55c303ad8. Root independently verified the logs, snapshots, 35 Package files and actual Wasm identity. Earlier caller and control failures remain preserved; no full-hour acceptance is claimed.
• Independent assessment accounts for every microsecond of interval 373, a 231.727452-second Quest 18 gap for Mage 1000070. Defense consumes 119.772839 seconds, with melee attacks and repeated changes to the latest attacker. The Character dies three times. Source review finds that Defense bypasses the existing class combat strategy. Four nested cast-finish OutOfRange Refusals are retained as a separate observation. This is a failed Quest defense cycle, not an unexplained Grind interval. Assessment SHA-256 fb5011732a38c45bd1a1af8dabcd34f6e51bd4ae15c4fcac4f1c3e64117e1dd8; machine account a77cab53c73d0ee197ce5b30c7bcb0ed1d864600c03b45763ce4db0de3235e6b. Defensive rotation, target retention and the remaining open intervals require correction or causal assessment before another hour.
• The final v16 interval ledger contains each index 0 through 557 exactly once. Eight open Quest 33 gaps repeatedly wait at a Completed destination, where the old expiry path never applies. Movement refreshes the navigation deadline and previously reset the progress clock. The correction uses 120 seconds since an actual Quest effect, independent of the navigation deadline, then the existing 30-second destination deferral. It retains the Quest and permits alternate work. Ledger SHA-256 8454415204fa05a743d97f1c7b8da6eb02792e7d338c7561970aa94bab00b65b; causal assessment a1b0ddc639dafd723467b65f6829e879ca53aef37459b9d941cef3244982c0f8. Classified contention remains an observed contributor, not proof that every long interval is acceptable.
• Offline behavior verifier v10 now counts nested cast-finish Refusals. The retained hour still has zero Refusal-spam violations and exactly the same 558 flagged intervals; 31 previously omitted nested Refusals enter the count. All other behavior fields are unchanged. Output SHA-256 1372dbe16650e3a97a98113d42afab3e2da88f85b2d3373fd11ab555e92683d4; transition receipt 97f63a995b97b64b1477e3b7069d8ecee72705287f48da45eeb7fd071e021589. Original capture, input manifest and earlier analysis remain unchanged.
• The new Mage Defense regression passes once on Core caller 76271989 and Collection 32a90a13. It proves real Fireball damage, retention of the current valid attacker after a second attacker hits, replacement after control or death, and retained Quest purpose. Exact gameplay log SHA-256 e7adce06696878137bf13a17f457f2b8c15e12d76288410d14e5ec1c9a6ba57d; actual Wasm 9ab407863bdd890b9213ab9500f7fa410ae05a0459fedf6a6046fc322feb7185. The v3 wrapper misread the split libtest output; the shared exact-pass parser and root independently confirm one pass. That original incomplete wrapper result remains unchanged. The old Defense boundary is being run separately.
• Core caller correction 2e6ed682 replaces fixed primary-plus-one expectations with the actual live Quest replacement and matching Recovery Fight. It also captures the Priest Smite completion transaction before later casts overwrite the current Action row. Standards and Spec review is clear; focused runtime checks remain queued. Diagnosis SHA-256 6ed7b16556cf3ad0d909a050a427d4ba85b40c1a3f15d8693934a74e3f45c0c4. These are test-only changes and do not clear the failed imported hour.
• Defensive combat focused checks are complete at Core caller 76271989 and Collection 32a90a13. The Mage regression passes in 119.26 seconds, and the unchanged progress and return-home boundary passes in 128.87 seconds. The final combined result is complete with successful cleanup. Result SHA-256 067173924c90a63a9c6a2d93e74c3b2961e0621198b283199875a28a3cf793cc; independent root review 716e4595d2aa3cd06265cafec1012adbb1c829b40010147d18d15da6d9c035eb. This closes the earlier incomplete wrapper and pending boundary work without replacing their original records.
• Core #519 at 97952c32 combines the reviewed collection, Defense, moving-cast and Completed Quest wait callers with the five CI observation corrections. Source review is clear. Formatting, Clippy, Rust tests, Wasm and Runtime Scripts checks pass; the Core durable check is still running. Core production code is unchanged. The Package durable suite includes all five corrected callers without a name filter; its next run awaits the final composed Package fixes.
• Completed Quest liveness passes both focused cases at Core 871b63cddc0ca893d21322936a10ffe161c33f74 and Collection e1be33aed368841212ba9411aa6ffffbe1ff5b5e. Real movement leaves the Quest effect clock unchanged. The existing two-minute bound defers the blocked Quest 7 despite a future navigation deadline, preserves the unrelated retry boundary and selects available Quest 5261 work. Foreign-tag damage leaves the clock unchanged; eligible damage advances it while the primary stays alive at 90 health with zero Quest credit. Exact test durations 15.56 and 11.82 seconds. Result SHA-256 2df1b786950500707905db9d3a9e6cef73e5c60e9b3fcf703930e5150424f32d; actual Wasm 1044fd89c22f3406e6e04b287d00a2ed8448098c7df6c0e4661ace82ff5f7f19. Root verified 39 Package inventory files and 9 bound build, log and snapshot files, review dadc8ca19b213a1e6d72278153b33b3a0366774979f7e6e6e84f1c350963625d. The evidence directory retains an older descriptive label; its result and build manifest bind the exact tested source above.
• Moving-target casting passes its focused real-gameplay regression in 21.023 seconds. The Mage first moves into casting position, then resolves Fireball against the retained Quest 7 target and earns credit without an OutOfRange Refusal. Result SHA-256 382951ffba285164f5e21cacded192910ef15c1563aaded97495f3dc2c1a02e4; independent root review b1dbb17930ad07b2c03407653f107e6657ac1ddefc5e81471bfc29a0477abecd. Core 916bac3a gives the common state capture a distinct filename so final CI retains it alongside the custom diagnostic.
• The combined candidate is Core 916bac3a0649e9abc6030e7f3cd56816b83fc72c, tree 5496a3f533b6050a42757708cb1ad25f283c74f3, and Collection 4e8d98e3d29c8327b55ca679672a8386604620b2, tree d16b13959a2da85b3740ae69c929e2f342cbbeb4. Independent Standards and Spec reviews are clear. The recorded build and cleanup pass: result SHA-256 19431313a311db324dc8528b1404036833bb5f1d0e8df96462700719e803ffcd; build manifest ebdc8b1a9a80162cdcc20604ee3ed1c787bf30c03247acc7070556a4c3e95036; Wasm 64914c8d36ad1f7f251a580a2cc6515903e605bfb048688333661830bca61b0a, 9,690,011 bytes. All 39 retained Package inputs match the build manifest. The 25-bot imported hour started preparation at 20:47 UTC on 12 September, alongside current-head Package CI. Input manifest SHA-256 7c75f189abf34a23f8f9130cef31e25771969aaeb215f9af3162865a0caa4ac0. This records execution in progress, with no acceptance result yet.
• The running hour retains actual Collection 4e8d98e3 and Wasm 64914c8d. Delivery Collection 004dda30034a3d9668c482d0c6322934a2f94e24 changes only two test fixture reads to the existing map index and exact partition filter. Neither changed reducer is called during the imported hour. Production sources are byte-identical. Root source-equivalence receipt SHA-256 401a91d1dbfff4c32a2fb0fbb31d1585a48732d6c594ffd38282fcf9f9f54ec4. This does not claim identical compiled Wasm or a new accepted observation.
• The measured 3,600-second observation started at 20:59 UTC after all imports and setup checks passed. Core #519 now passes all required checks. Final Collection 004dda30 passes the module, Package API and Runtime Script jobs; Transfer, composition, durable and class acceptance continue. Both Core and Collection merges remain held for their declared evidence gates.
• The completed 25-bot imported hour retains Core 916bac3a, Collection 4e8d98e3 and Wasm 64914c8d. Its 3,600 measured seconds and private cleanup closed at 22:00:25 UTC. All bots reached level 4 or 5. Behavior analysis flagged 238 intervals, down from 558 in the preceding failed run. Ownership assessment covers all 238, including one unfinished interval; full causal review remains open. Run SHA-256 8e8814d01896161826735d155abad3967bf786460733c00110c88fbd37a5a49c; behavior b17be9d83ab3717a7f5517c0ceeea1ed6cc48c538637ba470e44376eed6dab3d. This hour is not accepted.
• Static geometry assessment checked 153,370 movement segments with zero blocked segments, incomplete paths or unexplained discontinuities. It retained 507 NPCs and accounted for 91 Release Spirit relocations. Geometry report SHA-256 b23ccd3d1672cc4ebaeb6e7be6ba5af45dd9d6baed8c406786bb3543948facc9. Rendered movement and dynamic client behavior remain separate human observations.
• Root confirmed all three Priest burst candidates as a Recovery selection defect. Bots 1000009, 1000061 and 1000165 record 10, 8 and 8 Refusals within 30 seconds. Every event is a Recovery self-cast of Lesser Heal 2050 receiving InsufficientPower or Cooldown. The candidate can cancel useful Defense movement or a pending Smite before Core refuses the heal. Current Core readiness must be checked before selection, while preserving a matching pending heal. Diagnosis SHA-256 6414c46b73ed9acca4fc8b918f9f4f1817ce304604ebc458ff50bedf1b07a570; raw window extraction 809fc4de0d1bd5f98a5986229a92a738a38704552e11a489f870f0a22ffa3c25. The correction is underway; the original failed hour remains unchanged.
• Delivery Collection 6efe7aa7d7aa952b319de826231f1fd5fdbb2f47 changes only three test reducers since observed 4e8d98e3. Two use indexed partition reads. One asserts current movement evidence instead of expecting navigation to advance a Quest effect clock. Root equivalence receipt SHA-256 f0dd584e289710640229594387607bd54419d2cf3d0f1b896c74a030570df403. Canonical push run 34720181322 composition passes all 46 unique executions, including the corrected movement expiry case; every command and cleanup status is zero. Artifact audit SHA-256 dce1ce06fc4492f40ae457118d46ae5d39eb5f0097879db47a934c3b8e3dcd2a. Durable and class closure remain pending.
• Independent review accounts for all 238 flagged intervals without timeline gaps or overlap. No additional source defect is established beyond Recovery readiness. The one unfinished advancement span contains active Grind travel and casts toward targets that begin eligible; no closing progress is claimed. All stationary no-effect spans are below 114 seconds. Independent report SHA-256 b710a687d58afccc1fa435da7dddf5cae173dfeaae239fd5944c705c9bf212bb; per-index ledger dcd522d1055fc70aa8e132f1753925d7fcbfef8016fc651d2043bc4c71165c3e; root review 4c63f605f7ef96f214fb91660ff890325153269819aebf4ffa71d27fb54b7c16. Final retained rows confirm that all 25 bots earned rewards for Quests 783, 7, 5261, 33, 18 and 3903. Progress receipt SHA-256 b5b9fd7fcf92703321be1a4bf2d8349227c64d5f60d62db269db39dedc3f6f36. This measured result is not an added release requirement and does not accept the failed hour.
• Canonical Collection run 34720181322 and its PR run pass all nine jobs at Core 916bac3a and Collection 6efe7aa7. The durable job passes 169 cases, comprising six Gateway and 163 Module cases. All five corrected existing observations and all five new defect cases pass. Composition passes 46 cases, class journeys pass 30 cases plus the Warrior level-gap case, and Transfer passes 27 cases. Root independently checked the durable log, result groups, required case names and distinct movement observation files. Final CI audit SHA-256 7b4a4db3b9f2545961d3a146f0a1036085be0bac86e8cf087bd83d4b62dcf36d; root durable review 53b389652e5938f6d8faac7216a93301c45ccb4a31d23866d493798bd1c9c4e0. This closes the preceding delivery checks only; the new Recovery production change remains under focused test.
• Recovery diagnosis revision 2 corrects the scope description of the retained data. World Entity records do include power and max_power; the 26 flagged Refusals have 0 to 29 mana. Dedicated cooldown rows were not subscribed, so the recorded Core Refusal establishes Cooldown. The event counts and causal finding are unchanged. Revised diagnosis SHA-256 5e60926a18b6bbe81cd94906ed81599b06369830811c6841704788a937430307. Original diagnosis and failed-hour evidence remain preserved.
• The focused Recovery test first reproduces the defect on unchanged production: at 25 mana, old selection cancels Defense casting-position movement for Lesser Heal 2050, then records InsufficientPower, have 25 and need 30. Causal RED result SHA-256 6db62a73994c172f25aa0ad707f6c55d0b4f8588b5f1b6816737871b23cbf0a1. Corrected Core 262d6b7b and Collection e0ed910e pass the exact case in 22.64 seconds. The three scenarios preserve casting-position movement during insufficient mana, preserve melee movement during a real global cooldown, and retain pending Smite before a ready heal takes priority. Every scenario completes Lesser Heal with 50 actual healing. All eight retained files match their hashes. The original wrapper result remains preserved because its same-line parser missed the passing output; separate strict assessment 284ecd95ec8e4636f57a18de09b6cb0d04d3a962434f5fa62a51d6307800cdb5 accepts the actual one-case pass. Root review SHA-256 1f729e54415202a379ceb691fab822fb2661db9a3bf388b888cf0bb0b564e491. Independent Standards and Spec review is clear. No repeated runtime was needed for the parser discrepancy.
• The final combined build at Core 27d6f533 and Collection 54a7d39d passes source, build and cleanup checks. Wasm SHA-256 cfdf2aa51ebcce07e85607d4ec6dcf820c32255d76833cc6859366dce4495faa, 9,792,774 bytes; build manifest 43dea32545c7d06d7e790d3faba07a9cfb36e1cc3ae743901baf4439ab9f2363; build result df63a65133c259c10d2a157cef7c80ab3c58181642082a40a47a76acf21f530c. All 39 retained Package input files match the manifest. Delivery 0622ca5e changes only one unexercised debug fixture error message; exact diff SHA-256 ee77baa70f00aa06573362744ffad7bdaa5b67a42d1ad5ffcb9ef1f456f14575. This source equivalence does not relabel the recorded Wasm. The new 25-bot hour began preparation at 23:35 UTC with input manifest 4e4361e29ed53ba6486bbd0f550177810199fd7d15c25907d7bcba36948c6c0b. Driver, observer, content, geometry, timing and behavior criteria are unchanged. Final CI runs alongside the observation; no new imported acceptance is claimed.
• Final Collection CI at 0622ca5e and Core 27d6f533 passes in canonical run 34725842185 and duplicate PR run 34725845639. Durable passes 170 tests across 13 result groups, comprising 6 Gateway and 164 Module cases. All five corrected observation callers and six gameplay regressions, including Recovery readiness, pass. Composition passes 47, Transfer 27, all three starter classes ten journeys each, and Warrior one additional level-gap case. Final audit SHA-256 a91f71907eb165210e4857f80a331dd86df5a436a089c0431b7390ac60a4d853; independent raw-log and artifact review c364a60d83bebb952aff97a2fb9f61ff9da1aacd78f162c2b8a2191116365da9.
• The full imported v18 run at Core 27d6f533 and recorded Collection 54a7d39d fails its timing-evidence check. It records 7,173 measured passes and 90,577 decisions, with 18,464 timing samples missing before midnight and none missing afterward. Finalized transaction archive SHA-256 84afaec6fb7cca06d10d6937a3695f25a8bde109b6543f06b6235528d5c1ad22; pass archive 678dad076bfe10443366eb781b3ddb1acbb8142788770c61e31f46f079e43108. Failed run a6b8b51de2840c212d884f380ba303a070366f30e19bd30f65c6490b4c55994d; exact census 7b7743d7de0b4e1e5f25d5a2094d9b2cd318d784c994a3a2a49b31b751677c5a. Resource bounds and cleanup pass, but missing timings and skipped final geometry reads prevent acceptance. Original files remain unchanged.
• Recorder correction #21 merges as 4043a3c0 with the exact reviewed and tested tree be546b078377a792fa974d18a1371e2d5dd60c4f. Native result SHA-256 54dba47fba3ae5b5cbc2e2322f51af964eb3143b984cf8ee609583e41a5a8fc9 records 36 passing tests, format, strict Clippy, release and cleanup. Binary 5c6377ecd5a15ddef7e07cca0be9d831b65465feb29b9867cae1c34b9c5bd28c is 1,628,160 bytes. Independent source and artifact review 48388a61c67e3d3ace825916c7f542aea3853f0af0200261dc104fc4e9a3ec3a verifies all 44 retained files. New v19 input a18ed06c22bd744e5ec2707702d8d5cb624f92f795d016baae53711c08f8f997 differs from v18 only in the recorder identity. Both reviews pass; import preflight passed and measurement began on 13 September at 01:30:00 UTC. The exact window is retained and the live timing stream is recording. Expected closure is about 02:30 UTC. No result is claimed.
• The failed v18 trace receives a diagnostic review at both admissible measurement boundaries. All 280 flagged spans have complete clipped accounting: 205 no-effect spans, 68 advancement gaps and seven injured-healer spans. There are zero repeated-Refusal burst candidates and no additional gameplay source defect established. Two active Grind intervals remain right-censored open observations. Review SHA-256 d79fee2017c8159dabc19b0fac931d23865d82b78f028bc7747fa0aeb619daa3; causal ledger 2031098529d851be046bd5f487848c1e2d8adc6dc61cf4ad805c263d58496551. Diagnostic recovery cannot replace missing timing, final geometry or exact window evidence; v18 stays failed.
• V19 closes captured at 02:30:51 UTC on 13 September using the same frozen Core 27d6f533, recorded Collection 54a7d39d and Wasm cfdf2aa5. Its 7,179 measured passes contain 90,583 exact timing samples. Both archives finalize within their limits and private cleanup passes. All 25 level-one bots end at level 4 or 5. Run SHA-256 5fde8f19391b154f66e43135cbfae75aed5eea79a3f2a4c0e29e295beaf8d463; capture e63903056e87c7c531f864919b2cf246a5b396e59c894286873e4f3f289c385b. Behavior, geometry and ownership analyses all close successfully. There are 269 flagged intervals and zero Refusal-spam candidates. Static geometry is clear across 146,375 assessed segments. All imported and measured-capacity reviews pass; root acceptance SHA-256 e681419b347641418f4194c55f756ded100449405aa3c6a2d41a5d0b856deaa9.
• Final retirement preparation is held at Core ec578c40/tree e0fb880a and Collection 5ae91707/tree b8c91b6a. The replay preserves accepted Recovery behavior and callers. Core review a5ec9d4f8d10551997b0ef018cc9e10eefca8491e34ecd3d70cb52460b9fec75 and Collection review 3c8127ec7089682fc36311fd84eb45db46c5ad85deba2a6b971e92157cf00c90 are clear. These exact preparations have not been built or run. Their final checks and delivery follow attended acceptance.
• The user-supplied Operator handoff on 14 September reports installed Core 3d152a3c and both Official Packages at 63f82468. Preflight failed because the CLI mistook imported enums for table rows. The previous live artifacts remain running. Deployment checks, publish, replay, repair, restart and staging were skipped. Map 36 also lacks its required active vmap generation.
• CLI #59 merges as c1cc373db1a77e850367ef3ec0a2fe612cbccfdf. Core pins reviewed source c556eb990b7d71671fdd1babad6e309f34400cc9. The regression failed before the fix; 724 CLI tests and Rust 1.85 check pass. Offline generation from the accepted Wasm now yields 299 parsed tables and zero errors, instead of 290 and nine errors. Shim installation and safe argument validation pass. This proves the parser correction, not Argus deployment or client acceptance.
• The later 14 September Operator handoff reports successful deployment checks, four published Shards, Package replay and post-publish repairs at Core b2531b18 / Collection 63f82468. Service Reconciliation failed after Standalone restart on stderr mode normalization. The old Gateway remains running with schema mismatches. Its replacement, Map 36 import and staging remain pending. This records the supplied Operator result, not a new live acceptance run.
• CLI #60 merged as 6cda0918296ada5910f2b078058836cf27882247. Core pins reviewed d738373bf36f25598e148d1dbdeef75ed6c244d8. The verifier checks append mode and the running stderr file. The regression failed before the fix; 728 CLI tests, Rust 1.85 check, independent review, pinned shim installation and safe argument checks pass. Live reconciliation and Gateway replacement remain pending.

Current notes

• The corrected v19 hour is accepted. All 90,583 decisions have timings, static geometry is clear, and all 269 flagged intervals have complete causal accounting. Three active Grind intervals remain explicitly right-censored; target contention explains their changes. Independent capacity review accepts 25 bots for the measured setup.
• Recorder PR #21 is merged after independent source review, 36 tests, strict Clippy, release build and all remote checks passed. It retains logs throughout measurement and saves the exact time window before aggregation. The failed v18 observation remains preserved. Its diagnostic trace established no additional gameplay source defect.
• Core #519 merged as 35495170 and Collection #36 as 63f82468. Both merge trees equal their tested heads. The observed Module retains its actual Core 27d6f533 and Collection 54a7d39d build identities; final delivery differs only in one unexercised fixture error message.
• The latest 14 September Operator handoff reports Core b2531b18 / Collection 63f82468, all six deployment checks passed with 3,739 tests passed and 30 ignored, four Shards published, and replay and repairs complete. Standalone restarted. Service Reconciliation stopped because the CLI expected append:/path but systemd reports append. The old Gateway still reports schema mismatches; its replacement is built but has not started. Argus is not ready for client acceptance.
• Retirement source is prepared and reviewed at Core ec578c40 and Collection 5ae91707. It preserves accepted Recovery readiness and all current callers. The earlier native checks retain their actual source identities. Final checks and retirement delivery follow attended acceptance; stored Legacy meanings and old-table lifecycle handling remain preserved.
• Earlier failed observations and their original evidence remain preserved. The latest successful private cleanup did not need the preceding run's cleanup repair.

Correct the importer before release observations. Read each M2 collision section independently of rendering records, preserve valid obstacles, and stop an import when collision data is malformed. Replay the declared Northshire client inputs with a pinned importer.

Acceptance criteria

• The observed missing-collision defect has a reproducible generated fixture, and valid collision survives unrelated full-model parsing failures.
• The shared collision reader checks section bounds, complete triangle indices, valid vertex references and finite coordinates. Nav, vmap and GameObject imports propagate a collision failure; genuinely empty collision remains empty.
• A clean pinned importer build passes its relevant checks and replays the declared Northshire collision and navigation inputs without the six former model skips. Evidence names source and executable hashes, archive inputs and actual geometry counts.
• The Core correction has independent source review, passing required checks and a normal merged PR before release acceptance.

Verified criteria: 4/4.

The observed missing-collision defect has a reproducible generated fixture, and valid collision survives unrelated full-model parsing failures.

Reviewed by Independent quest agent source review; orchestrator pinned input, runtime evidence and merge verification.

• The generated invalid-rendering regression first reproduced missing collision, then passed with the header-only reader. The read-only six-model replay decoded 96, 0, 96, 12, 0 and 0 collision triangles.

The shared collision reader checks section bounds, complete triangle indices, valid vertex references and finite coordinates. Nav, vmap and GameObject imports propagate a collision failure; genuinely empty collision remains empty.

Reviewed by Independent quest agent source review; orchestrator pinned input, runtime evidence and merge verification.

• Independent review checked bounded sections, complete triangle indices, finite coordinates, valid vertex references and propagation through navigation, vmap and GameObject readers. Focused malformed and empty collision tests pass; all 364 importer tests pass with 8 ignored.

A clean pinned importer build passes its relevant checks and replays the declared Northshire collision and navigation inputs without the six former model skips. Evidence names source and executable hashes, archive inputs and actual geometry counts.

Reviewed by Independent quest agent source review; orchestrator pinned input, runtime evidence and merge verification.

• Clean Core 14cda70a replayed all nine hash-verified archives. Executable SHA-256 23ff3ec6e391deb6bdec584a5dc63553c356a3e5069792714446191ce6f7a28f. All 75 models across 785 placements parsed without failure. Placed model triangles increased from 14,348 to 14,660, and the 25-cell Northshire navigation check passed.

The Core correction has independent source review, passing required checks and a normal merged PR before release acceptance.

Reviewed by Independent quest agent source review; orchestrator pinned input, runtime evidence and merge verification.

• Independent source review passed. All required checks passed on the exact reviewed head, including the durable suite https://github.com/LyraCoreProject/LyraCore/actions/runs/34348730052 . PR #499 merged normally as 8934c2df9da3e2e017df7d7b19cc2b6c5999426f.

Execution model: gpt-6-astra, high effort. Repositories: LyraCore.

Recorded evidence

• Read-only inspection found valid collision in three models that the full-model parser skipped: hearse 96 triangles, wagon 96 and stone pyre 12. The other three affected models contain no collision triangles. No client data was added to source control.
• Core14cda70a adds the checked collision reader. A generated regression first reproduced an empty mesh where one triangle was required, then passed with the correction. Three focused checks pass. A read-only probe using this exact parser decoded the six local models as 96, 0, 96, 12, 0 and 0 triangles. Independent review is clear. Full pinned importer checks and whole-scene replay remain pending.
• The clean pinned Core14cda70a build passed all 364 importer tests with 8 ignored. All nine source archive hashes matched the recorded inputs. The complete tile parsed 75 M2 models across 785 placements with zero failures and recovered 312 placed collision triangles, increasing the count from 14,348 to 14,660. The 25-cell Northshire navigation check passed. The built executable SHA-256 is 23ff3ec6e391deb6bdec584a5dc63553c356a3e5069792714446191ce6f7a28f. No realm import or client gameplay ran. PR #499 is open for CI and review.
• Core #499 merged normally as 8934c2df9da3e2e017df7d7b19cc2b6c5999426f after every required check passed at the reviewed 14cda70a head. Independent source review and the pinned local collision and navigation replay are complete. The importer prerequisite is complete; imported-world and attended-client gameplay remain separate pending requirements.

Current notes

• This child ticket follows an observed input defect. It is required before the imported-world and attended-client observations; it does not replace either release gate.