docs(render): resolve #458 — the a9c9 doorway admission is a 0.5 % edge-plane precision boundary (AD-118)

Two live cdb captures on the paired 2013 client at the Holtburg doorway:
the blockset template (every Render::block_check call of one frame — 2,601
resident blocks x 2 exit views) proves retail tests a9c9 at ring slot
25,46 and returns PARTIALLY then OUTSIDE; the blockcheck template dumps
its four corner interval vectors (0 0 0 0 300.4|310.2 and 0 0 0 1001 ...,
slab 75..330). acdream's replay at the same P pose reproduces the sentinel
pattern, the four edge planes and both verdicts, drawing a9c9 once, with
clip heights 298.8/308.5 m — a 0.5 % plane difference (about 0.35 px of
projected door-vertex position). At the fixture pose that margin is what
flips the south-west corner from inside to outside on the fourth edge
plane, so retail's four-corner unanimity test says OUTSIDE where ours
says PartiallyInside. Not fixable bit-exactly short of D3D's x87
transform; the KnownFailure row stays with its comment rewritten, the
register carries AD-118, and the first filtered capture's silent miss
(cdb sign-extends poi() inside .if; use dwo()) is noted in the template.

Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>

docs/ISSUES.md

@@ -41,7 +41,7 @@ remain closed. See
 
 ## #458 — Walk admits land block `a9c9` at holtburg-doorway-still that retail does not draw
 
-**Status:** OPEN — found 2026-09-03 by the S3 chunk 1 eight-kind transcript conformance (the gate's first real walk catch).
+**Status:** RESOLVED 2026-09-03 — a PRECISION BOUNDARY, not a walk defect (register row AD-118; the KnownFailure row is kept on purpose — see the resolution below). Found 2026-09-03 by the S3 chunk 1 eight-kind transcript conformance (the gate's first real walk catch).
 **Severity:** MEDIUM (retail-order exactness; one far land cell drawn that retail skips — no known visible artifact)
 **Component:** `WalkLandscape.CheckBlocks` / `LandCellCheck` block-level visibility at a LOD-ring boundary
 
@@ -99,6 +99,32 @@ pose against the decomp (the per-view union, the block MinZ/MaxZ clip heights, t
 exit-view edge planes, not the cell test. Resolve in the S3 review round or an S3 chunk 5;
 remove the KnownFailure trait in the same commit.
 
+**Resolution (2026-09-03 afternoon, lead + owner, two live cdb captures on the paired 2013
+client at a NEARBY doorway pose — `docs/research/2026-09-01-overhaul/oh-capture/
+holtburg-doorway-still.blockset.log` and `.blockcheck.log`, templates
+`tools/walk-oracle/oh/oh-capture-blockset.cdb.template` / `oh-capture-blockcheck.cdb.template`):**
+(a) Retail DOES test `a9c9` — `draw_check_blocks` calls `Render::block_check` for ring slot
+gx=25 gy=46 (`did=a9c9ffff`) twice a frame, once per surviving exit view (2,601 resident
+blocks × 2 views = 5,202 calls + 477 from the second caller @0x00505243). The first
+filtered capture printed nothing because cdb sign-extends a 32-bit `poi()` inside `.if`;
+`dwo()` fixed it (template note). (b) At the afternoon pose retail returned PARTIALLY
+(1) in view 1 and OUTSIDE (0) in view 2 with the four corner interval vectors
+`bound[0..4] = 0 0 0 0 300.4|310.2` (view 1) and `0 0 0 1001 300.4|310.2` (view 2), z slab
+75..330 — and acdream's replay at the SAME `P` pose (a throwaway test, deleted) produced
+the SAME sentinel pattern, the SAME four edge planes, the SAME verdicts (1 then 0) and ONE
+`LC:a9c9` token, with clip heights 298.81/308.49 m vs retail's 300.4/310.2 m: a 0.5 %
+difference in the doorway's fourth edge plane (its slope ratio 0.05042 vs 0.05104 — about
+0.35 px of projected door-vertex position at 720p). (c) At the FIXTURE pose (frame 2 of the
+morning walk capture) acdream's replay has that fourth plane classify three block corners
+OUTSIDE (1001) and the south-west corner INSIDE (0): `block_plane_check` → PartiallyInside
+→ the block draws. Retail's plane, 0.5 % away, puts the last corner outside too → OUTSIDE →
+no `LC`. One far LOD block, one corner, one plane, at a boundary. The morning z-range and
+structural notes above stand (slab and loop shape match). No bit-exact fix exists short of
+reproducing D3D's x87 projection and `copy_view`'s float rounding; the conformance row stays
+`Status=KnownFailure` with its doc comment rewritten to this finding, and the register
+carries it as AD-118. The two look-in-flood divergences (2)/(3) above are the pre-existing
+FW1 near-win gate state and remain tracked by `WalkLookInGateSweepTests` (skipped).
+
 ## #457 — Two App tests pass only in Release (IL-offset pins fail under Debug)
 
 **Status:** OPEN — found 2026-09-03 during the S3 chunk 3 re-review; pre-existing at `dffd7055` (before chunk 3), not a campaign regression.