diff --git a/docs/ISSUES.md b/docs/ISSUES.md
index 2dce41a7..d1e2de8b 100644
--- a/docs/ISSUES.md
+++ b/docs/ISSUES.md
@@ -589,13 +589,38 @@ the delta-review round on the same route's remediation. Evidence:
 
 ## #309 — Cancelled lost-cell park re-shows the entity where retail would keep it hidden
 
-**Status:** OPEN — but **largely superseded by #312** (closed `b1f914d5`,
-2026-08-04). #312 made a cancelled park restore the presentation half as well
-as the Runtime half, which is the behaviour #309's connected check was written
-to probe. What remains genuinely open here is the narrower retail-faithfulness
-question: retail's `GotoLostCell` keeps a lost-cell object HIDDEN until
-`reenter_visibility` fires on cell arrival, whereas acdream re-shows it on the
-cancel. Re-scope before running; the original six-step gate is now partly
+**Status:** **DEFERRED as an ACCEPTED DIVERGENCE — user decision 2026-08-06.**
+The standing record is register row **AP-136**, which already carries the
+retail mechanism, the exact divergence, and the observable; this issue is no
+longer a planned fix and does not block the placement-cutover campaign or C5c.
+
+Decision and its reasoning, so a successor does not silently re-litigate it:
+the retail-faithful end state is a park that SURVIVES cancellation, and that
+was implemented and reverted this round. Landing it costs (a) reversing a
+deliberate shipped invariant — `NewerPositionPickupAndParentEachCancelExactLostOperation`
+asserts that a newer Position cancels the park — and (b) `GameRuntime`
+teardown convergence (stage 10), where surviving parks never converge on
+shutdown. Weighed against an observable that requires a remote to teleport
+into a non-resident landblock **and then stop moving** (ACE stops
+broadcasting for a stationary entity; the ordinary 5–10 Hz case is
+superseded within ~150 ms), the cost is not currently worth paying. Revisit
+if the teardown-convergence work is done for another reason, or if the
+observable is reported in ordinary play.
+
+**This deferral does NOT cancel the six-step connected check.** That check
+validates the SHIPPED rollback behaviour (#312 / AP-136's `restorableOnCancel`
+path, which sits in `SubmitPreparedPlacementCore` — the shared core behind
+every production placement), not the deferred fix. It still needs running with
+`ACDREAM_PROBE_PARK=1`, and therefore must run **before** the C5c probe strip
+retires that flag.
+
+Prior status, retained for context: **largely superseded by #312** (closed
+`b1f914d5`, 2026-08-04). #312 made a cancelled park restore the presentation
+half as well as the Runtime half, which is the behaviour #309's connected check
+was written to probe. What remained genuinely open was the narrower
+retail-faithfulness question: retail's `GotoLostCell` keeps a lost-cell object
+HIDDEN until `reenter_visibility` fires on cell arrival, whereas acdream
+re-shows it on the cancel. Re-scope before running; the original six-step gate is now partly
 redundant.
 **Severity:** MEDIUM
 **Filed:** 2026-08-04
diff --git a/docs/architecture/retail-divergence-register.md b/docs/architecture/retail-divergence-register.md
index 4e263688..a4b8b30a 100644
--- a/docs/architecture/retail-divergence-register.md
+++ b/docs/architecture/retail-divergence-register.md
@@ -294,7 +294,7 @@ AP-94..AP-112 for the confirmed retail-UI completion gaps.
 | AP-133 | **Filed 2026-08-03 (#282).** A retail `CPhysicsObj` has exactly ONE `cell`; `ShouldDrawParticles` @0x0050fe60 reads that same field and calls `IsInView` on it, and `set_cell_id` @0x0050f4f0 / `change_cell` @0x00513390 are the only things that move it. acdream splits the concept into `WorldEntity.ParentCellId` (render parent, null for outdoor dat stabs and building shells) and `WorldEntity.EffectCellId` (authored landcell for those parentless stabs). Every consumer now resolves through the single `WorldEntity.VisibilityCellId` accessor (`ParentCellId ?? EffectCellId`); live entities carry `ParentCellId` only. | `src/AcDream.Core/World/WorldEntity.cs` (`VisibilityCellId`); writers `LandblockLoader.cs:80,97`, `LandblockBuildFactory.cs:408` | Outdoor dat stabs deliberately keep a null render parent so portal visibility does not filter them as interior geometry, yet retail still gives their physics object a landcell for particle gating. One accessor keeps the two fields from being read in conflicting orders, which is exactly how #282 arose - `EntityEffectPoseRegistry` preferred `EffectCellId` while `WbDrawDispatcher` and the remote spawn seed preferred `ParentCellId`. | A future writer that sets `EffectCellId` on a live entity re-creates #282: it wins `VisibilityCellId` while the 11 per-tick `ParentCellId` writers leave it frozen, stranding that entity's particles and lights on a stale cell so they fail `IsInView` after it crosses a boundary. | `CPhysicsObj::ShouldDrawParticles` 0x0050fe60; `CPhysicsObj::set_cell_id` 0x0050f4f0; `CPhysicsObj::change_cell` 0x00513390 |
 | AP-134 | **Filed 2026-08-03 (#297).** Retail keeps ONE `PublicWeenieDesc::_bitfield` per object and mutates it in place — `SetPlayerKillerStatus` @0x005AC7C0 rewrites bits 5/21/25 (PK `0x20` / Free `0x200000` / PKLite `0x2000000`, mutually exclusive), driven from `ACCWeenieObject::OnStatUpdated` @0x0058DF20 `case 0x86`, and `IsPK`/`IsImpenetrable`/`IsPKLite` @0x0058C8xx read that same field. acdream replicates the value into FIVE stores: `ClientObject.PublicWeenieBitfield` (the source, written only by `ClientObjectTable.UpdateIntProperty` on PropertyInt 134), `InboundPhysicsStateController._snapshots[guid].ObjectDescriptionFlags`, `RuntimeEntityRecord.Snapshot.ObjectDescriptionFlags`, the decoded `ShadowObjectRegistry` registration + per-cell `ShadowEntry.Flags`, and the local player's `RuntimeMovementSkillState` own-PWD bitfield. Coherence is maintained by two `ObjectUpdated` subscribers (`RuntimeEntityPvpBitfieldSnapshotSync` for the two snapshot stores, `LiveEntityPvpBitfieldSync` for the decoded shadow flags) plus the appearance-rebuild path re-deriving from the snapshot. The two shadow-flag writers are the SAME invalidation applied at the two edges that can invalidate it, not competing authorities. | `src/AcDream.Runtime/Entities/RuntimeEntityPvpBitfieldSnapshotSync.cs`; `src/AcDream.App/Physics/LiveEntityPvpBitfieldSync.cs`; source writer `src/AcDream.Core/Items/ClientObjectTable.cs` (`UpdateIntProperty`, PropertyInt 134); decode `EntityCollisionFlagsExt.FromPwdBitfield` | ACE never re-sends a `PublicWeenieDesc` after login (`EnqueueBroadcastUpdateObject` has zero live callers), so PropertyInt 134 over 0x02CE/0x02CD is the ONLY signal a PK status changed — a client cannot learn it from the bitfield itself. The replication exists because acdream separates wire snapshots, canonical records, and the collision shadow registry, which retail does not; each layer needs the decoded value at a different lifetime. Before #297 the snapshot stores were immutable wire captures; this commit is what converts them into write-through caches, and therefore what creates the invariant. | Any future write path that sets `ClientObject.PublicWeenieBitfield` outside `UpdateIntProperty`, or any NEW decoded cache of the PK bits, silently re-creates #297: the player walks through PKLite opponents and melee/missile admission refuses them, with no test failing. Note the same class already exists one field over — `Properties.Ints[134]` is written by `UpsertProperties` (PlayerDescription 0x0013) and `UpdateProperties` (IdentifyObjectResponse) WITHOUT mirroring into the bitfield (#300), and retail's `OnStatUpdated` also rewrites `_blipColor` (`case 0x5f`) and `_radar_enum` (`case 0x85`) which acdream ignores entirely (#301). | `PublicWeenieDesc::SetPlayerKillerStatus` 0x005AC7C0; `ACCWeenieObject::OnStatUpdated` 0x0058DF20 (`case 0x86`); `ACCWeenieObject::IsPK`/`IsImpenetrable`/`IsPKLite` 0x0058C8xx; retail `PKStatusEnum` `acclient.h:6412-6427` |
 | AP-135 | **Filed 2026-08-03 (C4 route 4a).** Retail `CPhysicsObj::MoveOrTeleport` 0x00516330 writes NOTHING on the airborne no-op (`arg4 == 0` -> `return 0` @0x0051636D), and `SmartBox::HandleReceivedPosition` 0x00453FD0 skips `ConstrainTo` with it (@0x00454272 sits inside `if (MoveOrTeleport(...) != 0)` @0x00454254). acdream honours that for every retail-modeled write — body pose, interpolation queue, leash, render entity, collision shadow, and the AP-80 velocity-derived animation cycle — but deliberately KEEPS two acdream-only per-packet bookkeeping writes on that branch: `RemoteMotion.CellId = wire landblock` and the `LastServerPos`/`LastServerPosTime` sample. This was pre-existing player-remote behaviour; route 4a extends it to NPC remotes so both arms are identical | `src/AcDream.App/Physics/LiveEntityNetworkUpdateController.cs` (`OnPosition`, both remote airborne-no-op returns) | The cell id is what acdream's OWN per-tick free-fall `ResolveWithTransition` sweep gates on (`rm.CellId != 0`); without it an airborne remote's sphere sweep is skipped and it falls through the floor (#42's neighbourhood). The server sample is what the first grounded packet after the arc synthesizes its velocity from; dropping it would make that velocity span the whole jump. Neither is a retail `CPhysicsObj` field being written | A remote's cell membership tracks the server's landblock during an arc where retail would keep the cell its own physics last resolved. Visible only if the server's mid-arc landblock disagrees with the client's swept cell — the wire cell is authoritative in every case acdream has observed. Retire together with the free-fall sweep gate, when the remote arc is resolved by the same transition machinery the local player uses | `CPhysicsObj::MoveOrTeleport` 0x00516330 (@0x0051636D `return 0`); `SmartBox::HandleReceivedPosition` 0x00453FD0 (@0x00454254/@0x00454272) |
-| AP-136 | **Filed 2026-08-04 (C4 route 4b-1 review). AMENDED 2026-08-04 (cancelled-park presentation rollback): this row's central claim — "the entity becomes VISIBLE IMMEDIATELY at the committed destination pose" — was true only of the CANONICAL half until that fix, and the gap was a defect, not a divergence.** `ParkDeferred` publishes a `Withdraw` receipt whose presentation half the host sink performs (graphical bucket, projection visibility, plugin world state/events, effect-pose registry, local-player shadow, selection), and `RestoreParkWithdrawal` cannot reach any of it. Its only mirror image was a LATER `Place`, which a remote that parks on its final Position and then stops moving never receives, because ACE stops broadcasting for a stationary entity — so the entity was left simulated, collidable, and audible while ABSENT from both the world render and the radar for the rest of the session. The rollback now publishes `RuntimePlacementProjectionKind.WithdrawalRestored` on the same ordered receipt stream, gated on the entity ending the rollback canonically whole (`FullCellId != 0` and `InWorld`) — deliberately NOT on this row's residency arm alone, which the shipped graphical remote path correctly skips because its per-packet prologue rebucket already recommitted a non-zero `FullCellId`. The AP-136 residual below is unchanged and is now actually observable. Selection alone is not re-established: see AD-63. Retail has NO cancel for a lost-cell park. `CPhysicsObj::SetPositionInternal` @0x00515BD0 commits the destination pose with `store_position` @0x00515CE2 and registers the object via `CObjectMaint::GotoLostCell` @0x00515CF2 (@0x00508210); the registration is removed by exactly one thing, `CObjectMaint::InitObjCell` @0x00508260, which drains the lost list on cell load and calls `CPhysicsObj::reenter_visibility` @0x00508296 (@0x00516250) to re-place at the committed pose. An update that performs no SetPosition leaves the registration untouched, so retail keeps the object HIDDEN until its cell loads. acdream's accepted-Position merge cancels the park instead (a shipped, tested invariant), so on cancel we roll the withdrawal back — `InWorld`, object clock, canonical residency — and the entity becomes VISIBLE IMMEDIATELY at the committed destination pose, uncollidable until its landblock publishes. The pose itself is retail-exact and is deliberately not rolled back. The `ShadowObjectRegistry.Suspend` applied by `WithdrawCanonical` is also not lifted, because un-suspending needs a real placement dispatch (`ReplacePositionRows`); the entity rejoins the broadphase on its next placement | `src/AcDream.Runtime/Physics/RuntimeSetPositionState.cs` (`ParkDeferred`'s `restorableOnCancel`, `Forget`, `RestoreParkWithdrawal`) | The alternative — leaving the cancelled park's withdrawal in place — strands the entity invisible AND intangible for the rest of the session, because `CancelCoreDeferred` restores none of it and the operation that was the only thing able to wake it is gone. Restoring at the pre-park pose was tried and is wrong: retail commits the destination pose, and route 2's tests pin that the pose survives the cancel. Restore covers the plain unplaceable-destination park and — **narrowed 2026-08-04 at the C4 route 4b-2 delta review, then relocated at that slice's round 3** — `SubmitPreparedPlacementCore`'s two collision-prefix-QUIESCENCE parks. **Corrected round 4 (D1): NOT "unconditionally" for any of the three.** Since the relocation, the same post-snap quiescence test gates EVERY park including the plain one, which the row's own next sentences already described; the word contradicted them. The original blanket "no quiescence park is restorable" was over-broad: its stated reason — re-admitting a spatial root into a retiring prefix blocks the retirement — is exact for `ParkCollisionResidents`, where the entity's OWN cell is retiring, but `TryGetBlockingQuiescence` also fires on prefixes the placement merely TOUCHES (any `QueriedCellIds` entry, i.e. a NEIGHBOUR landblock the sweep crossed a seam into; and the request's `CurrentCellId`, which on a FIRST submit names the destination rather than the departed source because both accepted-Position callers commit the accepted wire cell to `record.FullCellId` before submitting — scoped at round 4 (D5): a retained retry re-submits with no fresh merge, and a non-Position rebucket writer (the projection materializer `DatLiveEntityProjectionMaterializer` — C4 route 4b-3 deleted the second shipped writer, `RemoteTeleportController`'s rollback, and C4 route 7 D4 demoted the third, the equipped-child renderer `EquippedChildRenderController.TickChild`, to a presentation-only bucket move that no longer touches `record.FullCellId`) can rebucket that field to a third landblock, so the arm is live). The decision is taken inside `ParkDeferred`, AFTER `SnapToCell`, as `!IsCollisionPrefixQuiescing(body.CellPosition.ObjCellId)` — the cell `RestoreParkWithdrawal` will actually restore residency into, tested against EVERY live quiescence rather than against the single minimum-`OperationId` token `TryGetBlockingQuiescence` happened to return, and read after `LandDefs.AdjustToOutside` may have moved it (the reachable half of that, and the one a test now pins, is the re-derived cell: a wire (cell, position) pair whose position lies past its own named block's seam is exactly the pair #107's re-derivation distrusts, and it lands residency in a NEIGHBOUR landblock — see `QuiescingOwnPrefix_SeamCrossingParkIsRestoredAtTheReDerivedNeighbourCell`). A cell id of 0 is `AdjustToOutside`'s map-edge failure sentinel, never landblock (0,0), so the map is not consulted with it (C3c-F3). **Round 4 (D6) added the same test at RESTORE time**, in `RestoreParkWithdrawal`'s residency arm: the park-time answer is a snapshot, and route 2's park is RETAINED until the next packet's merge-time `Forget` ~150 ms later, so a prefix clean when the park was taken can be quiescing when the rollback runs. The `InWorld`/transient/clock half is still restored unconditionally — it is per-entity simulation state, not a claim on any landblock's collision generation. So the rollback re-admits nothing into ANY quiescing prefix, at the moment it actually writes residency rather than only as of when the park was taken, while leaving these parks non-restorable stranded the entity `InWorld = false` / clock suspended / not a spatial root with the only operation able to wake it destroyed by its own next accepted Position. A retirement park (`ParkCollisionResidents`) is still never restored, and now says so explicitly rather than relying on a parameter default | A remote — or the LOCAL PLAYER, which traverses the same shared core through route 2 — that teleports into a non-resident landblock and then STOPS MOVING stays visible at the destination without collision, where retail would hide it and re-show it on cell load — ACE stops broadcasting for a stationary entity, so no later packet corrects it. At 5-10 Hz the ordinary case is superseded within ~150 ms. Retire by making the park SURVIVE cancellation (issue #309), which is blocked on re-deciding the newer-Position-cancels-the-park invariant pinned by `NewerPositionPickupAndParentEachCancelExactLostOperation` and on teardown convergence. **This row carries a user-observable change to shipped paths** — `restorableOnCancel: true` sits in `SubmitPreparedPlacementCore`, the shared core behind every production placement — so it needs the two-client connected check written up in #309 — **rewritten by this slice, not merely extended (round-4 D2 correction: this summary used to describe only the original three remote steps plus "route 2's corrections are unchanged", which no longer matches the issue's own body)**. #309 is now six steps run with `ACDREAM_PROBE_PARK=1`: the three original remote-park steps, plus a quiescing swept-NEIGHBOUR step and a quiescing-DESTINATION step that both exercise the LOCAL PLAYER through route 2 and both carry a stated `[park]`/`[park-restore]` confirmation signal (a quiescence window cannot be synchronised by hand, so without one the step passes while broken), plus the unchanged-ordinary-correction step. Step 5 also asks the tester to confirm the destination landblock's retirement still COMPLETES, and states correctly that the deliberately non-restored park does NOT recover on the next ordinary Position | `CPhysicsObj::SetPositionInternal` @0x00515BD0 (@0x00515C1D/@0x00515CDA/@0x00515CE2/@0x00515CF2/@0x00515CF7/@0x00515D07); `CObjectMaint::GotoLostCell` @0x00508210; `CObjectMaint::InitObjCell` @0x00508260 (@0x00508296); `CPhysicsObj::reenter_visibility` @0x00516250 |
+| AP-136 | **Filed 2026-08-04 (C4 route 4b-1 review). AMENDED 2026-08-04 (cancelled-park presentation rollback): this row's central claim — "the entity becomes VISIBLE IMMEDIATELY at the committed destination pose" — was true only of the CANONICAL half until that fix, and the gap was a defect, not a divergence.** `ParkDeferred` publishes a `Withdraw` receipt whose presentation half the host sink performs (graphical bucket, projection visibility, plugin world state/events, effect-pose registry, local-player shadow, selection), and `RestoreParkWithdrawal` cannot reach any of it. Its only mirror image was a LATER `Place`, which a remote that parks on its final Position and then stops moving never receives, because ACE stops broadcasting for a stationary entity — so the entity was left simulated, collidable, and audible while ABSENT from both the world render and the radar for the rest of the session. The rollback now publishes `RuntimePlacementProjectionKind.WithdrawalRestored` on the same ordered receipt stream, gated on the entity ending the rollback canonically whole (`FullCellId != 0` and `InWorld`) — deliberately NOT on this row's residency arm alone, which the shipped graphical remote path correctly skips because its per-packet prologue rebucket already recommitted a non-zero `FullCellId`. The AP-136 residual below is unchanged and is now actually observable. Selection alone is not re-established: see AD-63. Retail has NO cancel for a lost-cell park. `CPhysicsObj::SetPositionInternal` @0x00515BD0 commits the destination pose with `store_position` @0x00515CE2 and registers the object via `CObjectMaint::GotoLostCell` @0x00515CF2 (@0x00508210); the registration is removed by exactly one thing, `CObjectMaint::InitObjCell` @0x00508260, which drains the lost list on cell load and calls `CPhysicsObj::reenter_visibility` @0x00508296 (@0x00516250) to re-place at the committed pose. An update that performs no SetPosition leaves the registration untouched, so retail keeps the object HIDDEN until its cell loads. acdream's accepted-Position merge cancels the park instead (a shipped, tested invariant), so on cancel we roll the withdrawal back — `InWorld`, object clock, canonical residency — and the entity becomes VISIBLE IMMEDIATELY at the committed destination pose, uncollidable until its landblock publishes. The pose itself is retail-exact and is deliberately not rolled back. The `ShadowObjectRegistry.Suspend` applied by `WithdrawCanonical` is also not lifted, because un-suspending needs a real placement dispatch (`ReplacePositionRows`); the entity rejoins the broadphase on its next placement | `src/AcDream.Runtime/Physics/RuntimeSetPositionState.cs` (`ParkDeferred`'s `restorableOnCancel`, `Forget`, `RestoreParkWithdrawal`) | The alternative — leaving the cancelled park's withdrawal in place — strands the entity invisible AND intangible for the rest of the session, because `CancelCoreDeferred` restores none of it and the operation that was the only thing able to wake it is gone. Restoring at the pre-park pose was tried and is wrong: retail commits the destination pose, and route 2's tests pin that the pose survives the cancel. Restore covers the plain unplaceable-destination park and — **narrowed 2026-08-04 at the C4 route 4b-2 delta review, then relocated at that slice's round 3** — `SubmitPreparedPlacementCore`'s two collision-prefix-QUIESCENCE parks. **Corrected round 4 (D1): NOT "unconditionally" for any of the three.** Since the relocation, the same post-snap quiescence test gates EVERY park including the plain one, which the row's own next sentences already described; the word contradicted them. The original blanket "no quiescence park is restorable" was over-broad: its stated reason — re-admitting a spatial root into a retiring prefix blocks the retirement — is exact for `ParkCollisionResidents`, where the entity's OWN cell is retiring, but `TryGetBlockingQuiescence` also fires on prefixes the placement merely TOUCHES (any `QueriedCellIds` entry, i.e. a NEIGHBOUR landblock the sweep crossed a seam into; and the request's `CurrentCellId`, which on a FIRST submit names the destination rather than the departed source because both accepted-Position callers commit the accepted wire cell to `record.FullCellId` before submitting — scoped at round 4 (D5): a retained retry re-submits with no fresh merge, and a non-Position rebucket writer (the projection materializer `DatLiveEntityProjectionMaterializer` — C4 route 4b-3 deleted the second shipped writer, `RemoteTeleportController`'s rollback, and C4 route 7 D4 demoted the third, the equipped-child renderer `EquippedChildRenderController.TickChild`, to a presentation-only bucket move that no longer touches `record.FullCellId`) can rebucket that field to a third landblock, so the arm is live). The decision is taken inside `ParkDeferred`, AFTER `SnapToCell`, as `!IsCollisionPrefixQuiescing(body.CellPosition.ObjCellId)` — the cell `RestoreParkWithdrawal` will actually restore residency into, tested against EVERY live quiescence rather than against the single minimum-`OperationId` token `TryGetBlockingQuiescence` happened to return, and read after `LandDefs.AdjustToOutside` may have moved it (the reachable half of that, and the one a test now pins, is the re-derived cell: a wire (cell, position) pair whose position lies past its own named block's seam is exactly the pair #107's re-derivation distrusts, and it lands residency in a NEIGHBOUR landblock — see `QuiescingOwnPrefix_SeamCrossingParkIsRestoredAtTheReDerivedNeighbourCell`). A cell id of 0 is `AdjustToOutside`'s map-edge failure sentinel, never landblock (0,0), so the map is not consulted with it (C3c-F3). **Round 4 (D6) added the same test at RESTORE time**, in `RestoreParkWithdrawal`'s residency arm: the park-time answer is a snapshot, and route 2's park is RETAINED until the next packet's merge-time `Forget` ~150 ms later, so a prefix clean when the park was taken can be quiescing when the rollback runs. The `InWorld`/transient/clock half is still restored unconditionally — it is per-entity simulation state, not a claim on any landblock's collision generation. So the rollback re-admits nothing into ANY quiescing prefix, at the moment it actually writes residency rather than only as of when the park was taken, while leaving these parks non-restorable stranded the entity `InWorld = false` / clock suspended / not a spatial root with the only operation able to wake it destroyed by its own next accepted Position. A retirement park (`ParkCollisionResidents`) is still never restored, and now says so explicitly rather than relying on a parameter default | A remote — or the LOCAL PLAYER, which traverses the same shared core through route 2 — that teleports into a non-resident landblock and then STOPS MOVING stays visible at the destination without collision, where retail would hide it and re-show it on cell load — ACE stops broadcasting for a stationary entity, so no later packet corrects it. At 5-10 Hz the ordinary case is superseded within ~150 ms. Retire by making the park SURVIVE cancellation (issue #309), which is blocked on re-deciding the newer-Position-cancels-the-park invariant pinned by `NewerPositionPickupAndParentEachCancelExactLostOperation` and on teardown convergence. **ACCEPTED AS A STANDING DIVERGENCE 2026-08-06 (user decision): #309 is DEFERRED, not planned.** The retail-faithful survive-cancellation end state was implemented and reverted this round; landing it costs reversing that deliberate shipped invariant plus `GameRuntime` teardown convergence (stage 10, where surviving parks never converge on shutdown), and the observable requires a remote to teleport into a non-resident landblock AND then stop moving. This row is therefore the permanent record rather than a staging note — do not re-open the retirement without new evidence that the observable occurs in ordinary play, or unless the teardown-convergence work lands for another reason. **The deferral does not cancel this row's connected check**, which validates the SHIPPED rollback path, not the deferred fix. **This row carries a user-observable change to shipped paths** — `restorableOnCancel: true` sits in `SubmitPreparedPlacementCore`, the shared core behind every production placement — so it needs the two-client connected check written up in #309 — **rewritten by this slice, not merely extended (round-4 D2 correction: this summary used to describe only the original three remote steps plus "route 2's corrections are unchanged", which no longer matches the issue's own body)**. #309 is now six steps run with `ACDREAM_PROBE_PARK=1`: the three original remote-park steps, plus a quiescing swept-NEIGHBOUR step and a quiescing-DESTINATION step that both exercise the LOCAL PLAYER through route 2 and both carry a stated `[park]`/`[park-restore]` confirmation signal (a quiescence window cannot be synchronised by hand, so without one the step passes while broken), plus the unchanged-ordinary-correction step. Step 5 also asks the tester to confirm the destination landblock's retirement still COMPLETES, and states correctly that the deliberately non-restored park does NOT recover on the next ordinary Position | `CPhysicsObj::SetPositionInternal` @0x00515BD0 (@0x00515C1D/@0x00515CDA/@0x00515CE2/@0x00515CF2/@0x00515CF7/@0x00515D07); `CObjectMaint::GotoLostCell` @0x00508210; `CObjectMaint::InitObjCell` @0x00508260 (@0x00508296); `CPhysicsObj::reenter_visibility` @0x00516250 |
 | AP-137 | **Filed 2026-08-04 (C4 route 4b-2); rewritten 2026-08-04 at the dual Opus review; REWRITTEN AGAIN 2026-08-04 (C4 route 4b-3) — the cell-less enqueue-vs-place delta this row existed to record is RETIRED, not merely re-scoped: the teleport arm now ports retail's `teleport_hook` verbatim and places unconditionally through the canonical Runtime placement owner, exactly like retail's `this_1->cell == 0` @0x00516386 branch. What survives is the two acdream-only divergences retail has no state for at all.** acdream can classify a remote's accepted Position into two states retail cannot reach, sharing ONE stated handler (`RuntimeRemoteFarSnapPosition.ResolveArm`'s `UnroutedCatchUp`, AP-87's shared `ApplyInterpolate` catch-up) instead of a duplicated near/far block. The states: (a) **no classification at all** — `RuntimeAcceptedPositionRouteRequests.TryBuild` refuses to fabricate a local-player position, so `ClassifyRemoteAcceptedPosition` returns null for EVERY remote packet until the local movement controller exists (the login window) and whenever the canonical record has not claimed a local id; **route 4b-3 adds two more null-producing reasons** — the merge observed no PRIOR canonical record for this entity, so the classifier has no honest pre-merge cell to feed the teleport predicate and declines rather than fabricate one (D1); and the dormant initial-residence enqueue path (`RuntimeEntityObjectLifetime.TryApplyPosition`'s `EnqueueDormant` return, reached BEFORE the method's own `PreMergeCommittedCellId` write), whose timestamps therefore always carry `PreMergeCommittedCellId: null` too — fix round 2026-08-04 (R8), unverified from static reading whether `OnPosition` reaches `ClassifyRemoteAcceptedPosition` for an enqueued packet at all, stated honestly rather than guessed; (b) **`RejectedAuthority`/`RejectedData`** — acdream validates wire authority and payload finiteness, retail validates neither. **D1 — the visibility arm is deleted, not merely narrowed.** Before 4b-3, `LiveEntityRuntime.TryApplyPosition` computed `projectionRequiresTeleportHook` as `pre-merge FullCellId == 0 OR !IsSpatiallyProjected OR !IsSpatiallyVisible` — a presentation predicate with NO retail analogue, since retail's `MoveOrTeleport` never reads visibility. That whole computation, the lifetime parameter, and the headless `false` argument are deleted; a not-visible remote's Position now classifies purely by distance/contact like any other, and visibility is presentation-only. **D2 — the wire-airborne leftover shape.** After the teleport/cell-less classification moves onto its own arm, a packet whose classification is null/`RejectedAuthority`/`RejectedData` AND whose wire contact bit is clear takes retail's return-0 shape: AP-135's two bookkeeping writes only (server-cell adopt, `LastServerPos`/`LastServerPosTime`), no body/queue/render write, no leash arm. This deletes the legacy player-arm fallback's entity-revert quirk (`entity.SetPosition(rmState.Body.Position)`) and unifies player and NPC remotes on one behaviour. **R3 (retained from the prior rewrite) — `RejectedData` is APPLIED anyway** when grounded. It is the one classification meaning "this payload failed validation" (`ClassifyAcceptedPosition` emits it for a `ValidPosition` failure and for a non-finite/negative derived `player_distance`), and `UnroutedCatchUp` hands the same payload to `ApplyInterpolate`. Not a regression — the legacy block did the same. **Headless (contract item 6) is satisfied vacuously and that is stated, not implied:** nothing in `AcDream.Headless` constructs `RuntimeRemotePlacementDriveController` (`SessionPlayerComposition` is the only construction site) and `RuntimeLiveEntitySessionController.OnPositionUpdated` returns early for every non-local GUID, so both the far snap and the teleport arm are graphical-host-only paths | `src/AcDream.Runtime/Physics/RuntimeRemoteFarSnapPosition.cs` (`ResolveArm`, `RuntimeRemoteAcceptedPositionArm.UnroutedCatchUp`); `src/AcDream.Runtime/Physics/RuntimeRemoteTeleportPosition.cs` (`OwnsTeleportPlacement`, the retired predicate's replacement); applied at `src/AcDream.App/Physics/LiveEntityNetworkUpdateController.cs` (`ApplyRemoteContactRouting`'s teleport check and default arm, the D2 wire-airborne shape); headless statement on `IRuntimeRemotePlacementServiceWindow` | AP-87's own snap conditions (`firstUp \|\| !willBeDrTicked \|\| bodyToTarget > 4 m`) still PLACE an unplaced or badly-lagging body for the two survivors, so a remote keeps tracking the server through the login window and through a rejected packet | The two survivors are unaffected by 4b-3: a leftover-classified remote beyond 96 m that is already tracking catches up over a packet interval instead of snapping — invisible in practice at that range. If AP-87's 4 m backstop were ever weakened, this arm would become a silent-freeze path | `CPhysicsObj::MoveOrTeleport` 0x00516330 (@0x00516386 cell-0/teleport — now ported, @0x005163AF near, @0x005163C1-E8 far); `CPhysicsObj::teleport_hook` @0x00514ED0; `RuntimeAcceptedPositionRouteRequests.TryBuild`; `GameRuntime.cs:288-290` (the no-fabricated-Vector3.Zero rule) |
 | AP-138 | **Filed 2026-08-04 (C4 route 4b-2, dual Opus review).** Retail's remote far snap is unconditional and unrefusable: `CPhysicsObj::MoveOrTeleport` @0x005163D9 calls `SetPositionSimple`, discards its `SetPositionError`, and returns 1 @0x005163E8, so `SmartBox::HandleReceivedPosition` arms `ConstrainTo` @0x00454272 every time. acdream's far snap is a canonical Runtime placement that can decline for reasons retail has no analogue for, and this row records the complete residual. **(1) An outcome that never reached the engine is a `store_position`; one that did is not.** Retail's `SetPositionInternal` @0x00515BD0 has exactly two shapes and acdream now represents both (**corrected 2026-08-04 at the delta review, which found the first version of this row asserting — wrongly — that no acdream non-commit outcome could represent the second**). STORES, because the resolve never ran: `Refused` (the pre-flight declined the destination), `Contention` (another authority owns the operation, or the Setup/world-frame preparation is retryable), `RejectedPreparation` (`RejectedAuthority`/`InvalidData` — preparation refused before anything was submitted), and `NotApplicable`. For those `ApplyAcceptedRemoteFarSnap` writes the accepted destination pose to the canonical body, exactly as retail commits it on the no-transition branch — `prepare_to_leave_visibility` @0x00515CDA, `store_position` @0x00515CE2, `GotoLostCell` @0x00515CF2, `return 0` @0x00515D07 — so the remote keeps tracking the server at 5-10 Hz, at the destination, with no resolved cell; retail would additionally have hidden it until cell load, which is AP-136's scope, not this one. DOES NOT STORE, because the resolve DID run and refused: `RejectedByPlacement` (`PhysicsEngine.SetPosition` returned a non-Ok error, acdream's port of retail's `CheckPositionInternal == 0` @0x00515C85/@0x00515CD5 and `curr_cell == 0` @0x00515C8F/@0x00515CB2, neither of which stores; or authority displaced after the engine ran, which includes the `CommitCanonical`-already-settled shape) and `Deferred` (Core parked, and `ParkDeferred` has ALREADY snapped the body to the parked result — the accepted destination for the pre-sweep park, the collision-settled `spherePath.CurPos` for the post-sweep one — which `RestoreParkWithdrawal` deliberately leaves alone). **(2) A quiescence park a far snap can provoke is now restorable at the source, not refused by a pre-flight.** **Rewritten 2026-08-04 at the delta review.** `CanAttemptDestination` (service window + Core's own `IsCollisionPrefixQuiescing`) reads ONE prefix, the destination's, and stays as an optimisation. It cannot be the correctness mechanism: Core's `PlacementTouchesPrefix` also matches the request's `CurrentCellId` (see the round-3 measurement below for what that arm actually names), and `ResultTouchesPrefix` scans every `QueriedCellIds` entry, a sweep footprint that spans NEIGHBOUR landblocks (`CellTransit.AddOutsideCell` re-derives the block id from the global lcoord and has no same-block filter) and does not EXIST until the sweep has run. Worse, the post-sweep check is `result.IsSuccessful && TryGetBlockingQuiescence(result, …)` and sits ahead of the restorable `result.IsDeferred` park, so a healthy about-to-COMMIT far snap near a seam was rewritten to `DeferredCell` and parked non-restorably. The fix is in `SubmitPreparedPlacementCore`: both quiescence parks are restorable, and `ParkDeferred` decides safety on the cell it will actually restore into — see AP-136 for the exact predicate and for why it does not re-open the retirement stall AP-136's blanket scoping was protecting against. On a FIRST submit the `CurrentCellId` half of `PlacementTouchesPrefix` is NOT the "source landblock a far snap is leaving": both accepted-Position callers committed the accepted wire cell to `record.FullCellId` before submitting (the graphical remote path through `LiveEntityRuntime.RebucketLiveEntity` in its shared prologue, route 2 through the merge), so that arm named the destination — measured 2026-08-04 at round 3. **AMENDED 2026-08-05 at the C5b architecture review: the route-2 half of that measurement is now STALE and the two callers no longer agree.** C5b made the merge withhold the wire cell (AD-60), and route 2 submits from `TryExecuteAcceptedLocalPosition` BEFORE the `OnPosition` prologue rebucket (W2) it returns ahead of — so on a route-2 FIRST submit `PlacementTouchesPrefix`'s `CurrentCellId` arm now names the SOURCE landblock the local player is leaving, not the destination. The graphical REMOTE half is unchanged: its prologue rebucket still runs ahead of the far-snap submit. The consequence is confined to which prefix the quiescence pre-flight matches, which this row's own part (2) already established cannot be the correctness mechanism (`SubmitPreparedPlacementCore`'s restorable parks are); it widens rather than narrows the set of prefixes a local force can be parked against. **Scoped at round 4 (D5): that is a first-submit property only, and the arm is live rather than dead code.** A RETAINED operation re-submits from its own cadence pump with no fresh merge (both drives re-read `record.FullCellId` at submit), and the surviving non-Position rebucket writer (the projection materializer — C4 route 4b-3 deleted the second shipped writer, `RemoteTeleportController`'s rollback, and C4 route 7 D4 demoted the third, the equipped-child renderer, to a presentation-only move that no longer touches `record.FullCellId`) can rebucket it to a third landblock, so a retry can genuinely name a third landblock — which `CanAttemptDestination`'s own doc already said and the two summaries elsewhere contradicted. **(3) The leash is not armed through a superseded incarnation.** Retail arms unconditionally on the nonzero return; acdream re-validates position ownership after the placement (the receipt is published synchronously and the projection sink can replace or delete the incarnation from inside it) and returns without arming if the owner moved. Both remote arms now run that check BEFORE their arming call — the player arm used to arm first, the NPC arm second, and one of the two mirror images had to be wrong | `src/AcDream.Runtime/Session/RuntimeRemotePlacementDriveController.cs` (`RuntimeRemotePlacementExecutionStatus` + `StoresAcceptedDestination`, `ApplyAcceptedRemoteFarSnap`, `StoreAcceptedDestinationPose`, `Advance`'s window-drop path, `CanAttemptDestination`, `SubmitAndResolve`); `src/AcDream.Runtime/Physics/RuntimeSetPositionState.cs` (`ParkDeferred`'s post-snap restorable decision and the two `SubmitPreparedPlacementCore` quiescence parks); `src/AcDream.App/Physics/LiveEntityNetworkUpdateController.cs` (both arms' re-validate-then-arm order) | The alternative to (1) is the shipped pre-review state: an emptied interpolation queue plus a stale body pose, i.e. a frozen remote that the next packet reproduces identically, since nothing about a refusal reason changes at packet cadence. That is strictly further from retail than either the deleted legacy block (which always tracked) or retail itself. The alternative tried and rejected in between — storing on EVERY non-commit outcome — is worse still in the other direction: it teleports the canonical body into a destination the engine's own sweep just refused, and overwrites a freshly settled pose (contact plane, step-down) whenever `CommitCanonical` landed and only the projection ownership was displaced. The alternative to (2) — keeping the pre-flight as the correctness mechanism and widening it — is structurally impossible, because the swept footprint half of Core's predicate does not exist until the sweep has run; the alternative of leaving the parks non-restorable strands the remote outright. The alternative to (3) — arming a leash on a host that is no longer the entity's canonical position owner — is a write through superseded state, the exact class the re-validation exists to prevent, and retail has no superseded-incarnation state for its unconditional arm to arbitrate | A remote whose destination this host cannot place into keeps moving and rendering but does not become collidable or cell-resident until a later packet commits — it can be walked through at range. Bounded by the 5-10 Hz packet stream and by how long the destination stays unpublished/quiescing. A remote whose destination the ENGINE refuses, or whose commit was displaced, keeps its last resolved pose for that packet instead of tracking — retail-exact, but it means a remote can look one packet stale near geometry it cannot be placed into. A quiescence park whose blocking prefix is a swept neighbour re-shows the entity immediately at the destination rather than hiding it until cell load (AP-136's own residual, now reachable through this path and through route 2's local-player corrections). **C4 route 4b-3 adds a second producer of the visible-without-collision shape in item (1)'s storing list**: the teleport arm inherits the identical store-and-stay-visible residual for the same reasons — a remote that teleports into a non-published landblock and stands still is visible but not collidable until a later packet commits. No new machinery; the retirement path is the same #309. A superseded incarnation's leash is left unarmed for one packet; the replacement incarnation arms its own on its next accepted Position. Retire (1) by making the far arm's failure path open retail's lost-cell registration instead of a bare pose write, which is issue #309's territory (the park must survive cancellation first) | `CPhysicsObj::MoveOrTeleport` 0x00516330 (@0x005163D9, @0x005163E8); `CPhysicsObj::SetPositionSimple` @0x005162B0 (flags `0x1012` @0x005162C4); `CPhysicsObj::SetPositionInternal` @0x00515BD0 (@0x00515C1D, @0x00515CDA, @0x00515CE2, @0x00515CF2, @0x00515CB2, @0x00515CD5, @0x00515D07); `SmartBox::HandleReceivedPosition` @0x00453FD0 (@0x00454254, @0x00454272) |
 | AP-139 | **Filed 2026-08-04 (Bug B).** The remote tick clears its InterpolationManager queue on the LANDING edge — retail’s own `set_on_walkable(1)` transition, the same edge HitGround fires from. Retail has no such clear on a ground or contact edge: its only queue teardown outside a completed walk is `PositionManager::StopInterpolating` from `CPhysicsObj::teleport_hook` @0x00514EFD and the `InterpolationManager::UseTime` @0x00555f20 stall/autonomy blips. The clear is carried over unchanged in intent from the deleted hand-rolled landing block (#184, 2026-07-07), which hung it on a hand-rolled `Airborne && IsOnGround && Velocity.Z <= 0` test that also fired on a steep (non-walkable) contact; Bug B re-derived the edge without changing the behaviour it was written for | `src/AcDream.Runtime/Physics/RuntimeRemotePhysicsUpdater.cs` (the SetPositionInternal commit block); the packet-side twin lives in `src/AcDream.App/Physics/LiveEntityNetworkUpdateController.cs` (`OnPosition`, the player-remote landing snap) | A contact-free arc never enqueues — route 4a's airborne no-op writes nothing at all — so anything still queued when the body lands is a pre-arc waypoint, and the first catch-up after touchdown would otherwise walk the body backward toward it | A remote that regains contact while a legitimately fresh waypoint is queued loses one correction and re-acquires it on the next accepted Position (~5-10 Hz). A body that repeatedly loses and regains contact (a bounce chain down a rough face) clears the queue once per bounce. Retire when the arc itself feeds the queue, at which point the pre-arc waypoints are no longer stale | `CPhysicsObj::teleport_hook @ 0x00514ED0` (`StopInterpolating` @0x00514EFD); `InterpolationManager::UseTime @ 0x00555f20`; `CPhysicsObj::SetPositionInternal @ 0x00515330` |