docs(phys L.2d): slice 1 + 1.5 shipped handoff + 3rd plan-of-record reframe

L.2d as scoped is essentially closed at the Holtburg site. The slice-1.5
trace settled the question: the "I can't walk through doorways" symptom
is a closed Door entity (Setup 0x020019FF named "Door") at the building
threshold, not a building-BSP-collision issue. Building BSP is healthy.

The two prior framings turned out wrong:
- L.2a handoff (2026-05-12): "per-cell walkability missing" — based on
  hit attribution pointing at the building, missed the Door cylinder
  also colliding per tick.
- L.2d slice 1 spec (2026-05-13 morning): "BSP shape fidelity, three
  hypotheses X/Y/Z" — ruled out by the trace once the probe labeling
  bug was fixed in slice 1.5.

Handoff doc captures full evidence, side findings (building double-
registration latent bug, missing PhysicsState in entity-source log),
and a candidates list for the next-session ordering discussion.

Plan-of-record L.2d sub-direction paragraph updated to match: "watch-
and-wait" mode, no more slices until a new shape-fidelity bug is
observed at a different site. Door-state handling becomes its own
sub-phase, scope deferred to project-ordering discussion.

Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>

docs/plans/2026-04-29-movement-collision-conformance.md

@@ -169,30 +169,40 @@ fallback.
 - Audit `Setup.Radius` and cylinder fallback behavior against retail before
   relying on them for conformance.
 
-Current sub-direction (revised 2026-05-13 in slice 1 design spec):
-The "I can't walk through doorways" symptom at Holtburg is **NOT a door-
-state-toggle issue** — the `[resolve]` probe captured 140 hit=yes lines
-at the doorway with `obj=0xA9B47900` (126 hits), one specific BSP shadow
-entry. The 2026-05-12 handoff initially proposed porting `CBuildingObj` +
-**per-cell walkability** as the fix, but reading
-[ACE BuildingObj.cs:39-52](../../references/ACE/Source/ACE.Server/Physics/Common/BuildingObj.cs)
-and named-retail
-[acclient_2013_pseudo_c.txt:701260](../research/named-retail/acclient_2013_pseudo_c.txt)
-shows that's **not how retail solves doorways**. `find_building_collisions`
-is just one BSP test on `PartArray.Parts[0]`. The doorway gap lives
-inside that part's physics BSP itself. Per-cell walkability
-(`CCellStruct::point_in_cell`, `sphere_intersects_cell`,
-`box_intersects_cell`, `CObjCell::find_cell_list`) is how the resolver
-selects **which cells** to iterate, not how it decides whether a wall
-has a hole — that work belongs to **L.2e**, not L.2d.
+Current sub-direction (revised 2026-05-13 evening after slice 1 + 1.5
+shipped and Holtburg-doorway capture analyzed — third reframe):
+L.2d as scoped ("shape fidelity: Sphere / CylSphere / Building Objects")
+is **essentially closed at the Holtburg site that motivated this phase**.
+Building BSP collision works correctly — the slice-1.5 probe captured
+real triangles in plausible world positions for `gfxObj=0x01000A2B` with
+`bspR=13.99m`. The 121 wall hits the L.2a probe attributed to
+`obj=0xA9B47900` were **side effects of the player already being pushed
+back by a separate Door cylinder entity** at the same doorway threshold.
 
-L.2d slice 1 is therefore a **read-only BSP-hit diagnostic** that
-captures full collision evidence per `[resolve]` `hit=yes` line.
-Distinguishes three hypotheses (wrong BSP loaded / over-registered
-parts / BSPQuery flaw) from a single Holtburg-doorway capture; slice
-2 is the right-sized fix scoped from slice 1's evidence. Design spec:
+The actual blocker is a server-spawned **Door** entity — Setup
+`0x020019FF` named `"Door"` — that ACE places at each Holtburg-town
+building threshold (five doors total observed across `0xA9B40029`,
+`0xA9B40154`, `0xA9B40155`). It registers as a Cylinder shadow entry
+via the server-spawn path; its Cylinder collision blocks the player
+walking into the doorway. That's **door-state handling**, a different
+class of problem from L.2d's shape-fidelity scope — it touches network
+(`CreateObject` PhysicsState bits), interaction (Use action on door
+entity), animation (door open/close), and collision-state-toggle.
+
+Recommend: **leave L.2d in "watch-and-wait" mode** with slice 1's probe
+infrastructure in place. No more L.2d slices until a NEW shape-fidelity
+bug is observed at a different site (dungeon walls, stairs, roofs) with
+the probe-armed client. The door-state work becomes its own sub-phase
+(probably nested under B.4 interaction or filed as a new L.2 sub-phase
+like L.2g) scoped separately.
+
+Full slice 1 + 1.5 handoff:
+[docs/research/2026-05-13-l2d-slice1-shipped-handoff.md](../research/2026-05-13-l2d-slice1-shipped-handoff.md).
+Design spec (now mostly historical, framing was wrong but probe
+infrastructure shipped from it):
 [docs/superpowers/specs/2026-05-13-l2d-cbuildingobj-collision-design.md](../superpowers/specs/2026-05-13-l2d-cbuildingobj-collision-design.md).
-Handoff: [docs/research/2026-05-12-l2a-shipped-l2d-handoff.md](../research/2026-05-12-l2a-shipped-l2d-handoff.md).
+Predecessor L.2a handoff:
+[docs/research/2026-05-12-l2a-shipped-l2d-handoff.md](../research/2026-05-12-l2a-shipped-l2d-handoff.md).
 
 ### L.2e - Cell Ownership: Outdoor Seams, CELLARRAY, cell_bsp