probe(physics): ACDREAM_PROBE_SUPPORT + ACDREAM_WIRE_MESH — separate #337's three candidates

The user is wedged at the top of Neftet rock plateaus, jumps sink into the
mesh, and a corpse falls straight through. ACDREAM_PROBE_REACH already ruled
out its own domain: blocked=0, every candidate tested-ok. Three candidates
remain — terrain support, a collision mesh not where its visual is, or the
transition wedging on an unobstructed path.

ACDREAM_PROBE_RESOLVE alone cannot separate them. It prints a three-value
contact-plane token, no plane normal, no plane height, no terrain sample and
no plane provenance, so all three produce the same line. Two additions:

[support] — one line per resolve for EVERY body, not just the player. A corpse
is a plain physics body with no player-specific logic, so its fall-through is
the cheapest available control on "movement code vs geometry data", and it is
invisible to any player-filtered probe. The line samples the outdoor terrain
INDEPENDENTLY at the body's own out-XY and prints the contact plane's own
height at that same XY. Two heights at one point make support=terrain /
object / none a measurement rather than an inference, and cpSrc= names the
site that asserted the plane so provenance and classification cross-check.

[geom] — once per GfxObj that comes near a mover: the object's physics-BSP
vertex cloud against its visual mesh AABB in the same local frame, through the
same prepared accessors the resolver queries. verdict=coincident REFUTES the
working hypothesis for that object outright; no-physics-bsp / empty-physics-bsp
/ displaced / extent-mismatch each name a specific data defect. Built to
refute, not to confirm — two diagnoses on this defect's lineage have already
been refuted by measurement.

ACDREAM_WIRE_MESH upgrades the existing F2 overlay, which drew a broadphase
proxy cylinder for BSP objects and so could not answer the question at all, to
the real physics-BSP polygon edges (cyan) beside the visual mesh box (magenta)
and the terrain surface (yellow). Own class per code-structure rule 1.

The provenance latch lives on PhysicsDiagnostics, not on CollisionInfo. Two
fields there first — the obvious home — broke the flat/graph differential
referee and the scratch-reset poison test, both of which compare CollisionInfo
member-for-member. Teaching either to skip a member is a one-line green fix
that puts a permanent hole in a referee whose whole job is comparing
everything. Captured as feedback_probe_state_off_compared_types.

Seven tests cover the support classifier's boundaries: a wrong classifier does
not fail to answer, it answers confidently wrong.

Gates: Release build 0 errors; complete suite 11,225 passed / 4 skipped / 0
failed from a cleaned tree — baseline 11,218/4/0 plus exactly the seven new
tests, skips unchanged.

Issue #337 filed with the symptom set, what is ruled out, and a table of what
each possible output means. All of this is TEMPORARY and recorded for
stripping with the physics-probe family.

Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>

docs/ISSUES.md

@@ -24,6 +24,91 @@ What does NOT go here:
 - Every session: scan OPEN issues at start; promote/close anything we touched during the session before ending.
 - Promoting to a Phase: mark as `DONE (promoted to Phase X)` + commit SHA where the Phase entry landed.
 
+## #337 — Neftet rock plateaus: wedged at the top, jumps sink into the mesh, corpses fall through — cause NOT yet established
+
+**Status:** OPEN — instrumented, not diagnosed. Awaiting the live capture below.
+**Severity:** HIGH — walk-through, fall-through, and a hard movement stop on world geometry.
+**Filed:** 2026-08-06, user-reported in live play after #334's fix landed.
+**Component:** physics / collision — possibly geometry data rather than movement code.
+
+### Symptoms, all from the user in live play
+
+1. Walks **up** a rock face onto a plateau fine, then **cannot pass at the top** — wedged, position frozen.
+2. Jumping is **"swallowed half way by the rock"** — the body sinks into the visual geometry.
+3. **A monster corpse falls straight through the rock.**
+
+Symptom 3 is the load-bearing one. A corpse is a plain physics body with no
+player-specific movement logic, so a fall-through there cannot be explained by
+anything in the player's controller.
+
+### What is already RULED OUT
+
+`ACDREAM_PROBE_REACH` (`334-fix-gate.log`, and the earlier
+`334-neftet-probe.log`). At the frozen position: `blocked=0`, and **every**
+candidate returns `tested-ok` — including the landblock's own rock mesh
+`gfx=0x010046DE` in cell `0x8766002B`. **No object is blocking the player.**
+That probe can only see shadow objects, so it has ruled out its own domain and
+can say nothing about terrain or the transition.
+
+Two diagnoses have already been refuted by measurement on this defect's
+lineage: the broadphase reach filter (#333/AP-158) and the edge-slide family.
+Do not open a third by reasoning from the source.
+
+### The remaining candidates — and what is NOT yet established
+
+- **(a) terrain** is what supports/blocks the body (walkable slope limit,
+  step-up refusal, terrain Z).
+- **(b)** a **collision mesh placed somewhere other than its visual**, so the
+  body interacts with geometry that is not where the rock is drawn.
+- **(c)** the **transition wedging** despite an unobstructed path.
+
+(b) is the current working hypothesis and is **NOT ESTABLISHED**. It is
+plausible — a corpse falling through and a jump sinking in are both what
+absent-or-displaced collision looks like — but no measurement supports it yet,
+and the instruments below were built to REFUTE it, not to confirm it.
+
+Possibly relevant, possibly coincidence: landblock `0x8766` carries the
+**largest single collision owner in the game**, an 81-cell (9×9) footprint
+measured during #334 — larger than anything else by a wide margin.
+
+### Instruments (2026-08-06 — TEMPORARY, strip with the physics-probe family)
+
+`ACDREAM_PROBE_RESOLVE` alone does **not** separate (a), (b) and (c): it prints
+a three-value contact-plane token, no plane normal, no plane height, no terrain
+sample and no plane provenance, so all three candidates produce the same line.
+Two additions close that:
+
+- **`ACDREAM_PROBE_SUPPORT=1`** → `[support]` + `[geom]`.
+  - `[support]`, one per resolve **per body** (players AND corpses): samples the
+    outdoor terrain independently at the body's own out-XY and prints the
+    contact plane's own height at that same XY. `support=terrain` /
+    `support=object` / `support=none` is then a measurement, not an inference,
+    and `cpSrc=` names the code site that wrote the plane so provenance and
+    classification cross-check each other.
+  - `[geom]`, once per GfxObj that comes near the mover: compares the object's
+    physics-BSP vertex cloud against its visual mesh AABB in the same local
+    frame. `verdict=coincident` **refutes (b)** for that object outright;
+    `no-physics-bsp` / `empty-physics-bsp` / `displaced` / `extent-mismatch`
+    each name a specific data defect.
+- **`ACDREAM_WIRE_MESH=1`** upgrades the existing F2 collision overlay from a
+  broadphase proxy cylinder to the objects' real physics-BSP polygon edges
+  (cyan) beside their visual mesh boxes (magenta) and the terrain surface
+  (yellow). Settles "visual versus collision" by eye.
+
+### How to read the capture
+
+| Observation | What it means |
+|---|---|
+| `[geom] verdict=no-physics-bsp` or `empty-physics-bsp` on the rock | The rock has **no collision geometry**. All three symptoms follow; nothing on the movement side needs explaining. |
+| `[geom] verdict=displaced` | **(b) confirmed.** Fix the placement/registration transform. |
+| `[geom] verdict=coincident` on every nearby object | **(b) refuted.** The cause is (a) or (c); read `[support]`. |
+| `[support] support=terrain` while standing on the visible plateau | (a): terrain, not the rock, is the support — terrain Z near the plateau top is the thing to look at. |
+| `[support] support=object` with `cpAboveTerr` ≈ the plateau height | The rock IS supporting the body; the wedge is (c). |
+| `[support] stalled=true ok=true` with `cpWalkable=true` | (c): the transition accepts the move and advances nothing. |
+| `[support] support=none` on the corpse throughout its fall | Nothing ever contacts it — consistent with absent collision, and `[geom]` says whose. |
+| **No `[support]` line at all** for the corpse's guid while it visibly falls | The client is not simulating that body — the descent is server-driven or presentational, and the client-side collision path is not the place to look. An absence here is a real answer, not a gap in the capture. |
+| `[support] cpWalkable=false` at the freeze | Slope-limit refusal — compare `cpNz` against `floorZ` on the same line. |
+
 ## #335 — The INDOOR half of retail's part-array `find_transit_cells` is not ported: an EnvCell neighbour is admitted on a SPHERE test where retail uses a BOX
 
 **Status:** OPEN