diff --git a/src/AcDream.App/Rendering/Walk/WalkVisibilityMath.cs b/src/AcDream.App/Rendering/Walk/WalkVisibilityMath.cs
index 9d5c0de2..3eb79c3c 100644
--- a/src/AcDream.App/Rendering/Walk/WalkVisibilityMath.cs
+++ b/src/AcDream.App/Rendering/Walk/WalkVisibilityMath.cs
@@ -232,17 +232,18 @@ public static class WalkVisibilityMath
     /// Id range: the full <c>0x0000</c>-<c>0xFFFF</c> landblock-prefix
     /// space, each landblock's cells enumerated <c>0x0100</c> through
     /// <c>0x0100 + LandBlockInfo.NumCells - 1</c> (the same range
-    /// <c>CellStructSurfaceConstructionInstalledDatTests</c>' own
-    /// independent OH2 walk uses — its pinned
-    /// <c>landblocksWithCells=3,405</c> diagnostic already agrees). What is
+    /// <c>CellStructSurfaceConstructionInstalledDatTests</c>' own OH2 walk
+    /// enumerates; that test reports its landblock count as a diagnostic
+    /// and does not pin the value). What is
     /// counted: one entry per <c>CellPortal</c> whose <c>PolygonId</c>
     /// resolves in its cell's <c>CellStruct.Polygons</c> AND has &gt;=3
     /// vertex ids that all resolve in <c>CellStruct.VertexArray</c> — an
     /// unresolvable <c>PolygonId</c>, a &lt;3-vertex polygon, or a cell
     /// whose own <c>EnvironmentId</c>/<c>CellStructure</c> does not resolve
     /// is skipped ENTIRELY (not counted toward the total or either plane
-    /// tally, matching this method's own live callers, which never see such
-    /// a polygon in the first place). Under that rule: 3,405 landblocks with
+    /// tally; no authored portal polygon has fewer than three vertices, so
+    /// the rule excludes only unresolvable data, none of which reaches this
+    /// method's live callers). Under that rule: 3,405 landblocks with
     /// cells, 1,854,237 portal polygons; 2,889 with at least one vertex on a
     /// +/-12 plane, 2,163 with EVERY vertex on the SAME plane — all 2,163
     /// are EXIT portals (<c>OtherCellId == 0xFFFF</c>; 0 interior portals
diff --git a/tests/AcDream.App.Tests/Rendering/Walk/WalkTraceConformanceTests.AlphaDepthTranscript.cs b/tests/AcDream.App.Tests/Rendering/Walk/WalkTraceConformanceTests.AlphaDepthTranscript.cs
index 85da59d1..ec628d78 100644
--- a/tests/AcDream.App.Tests/Rendering/Walk/WalkTraceConformanceTests.AlphaDepthTranscript.cs
+++ b/tests/AcDream.App.Tests/Rendering/Walk/WalkTraceConformanceTests.AlphaDepthTranscript.cs
@@ -341,10 +341,14 @@ public sealed partial class WalkTraceConformanceTests
     /// from terrace-edge's own first observed sample (<c>counterBefore=2</c>,
     /// carried in the file's pre-"F 1" preamble text — see
     /// <see cref="WalkAlphaDepthTrace.LoadInitialCounter"/>'s own doc
-    /// comment) closes fix round 1's KnownFailure: both of this fixture's
-    /// PM events now read the correct pre-capture value instead of a cold
-    /// <c>0</c>, and its PC sequence (<c>ov=0</c> throughout — this pose
-    /// never runs a qualifying interior turn at all) already matched.
+    /// comment) closes fix round 1's KnownFailure. Under a cold <c>0</c>
+    /// BOTH sequences diverged in the counter field — PM expected
+    /// <c>(1,2),(1,2)</c> and PC expected <c>(0,2,0),(0,2,0)</c> against a
+    /// cold <c>(1,0)</c>/<c>(0,0,0)</c> — round 1's write-up wrongly called
+    /// the PC side a match, and round 2's gate-honesty lens caught the same
+    /// sentence re-worded here (lead correction at landing). Seeded, both
+    /// sequences match exactly; this pose never runs a qualifying interior
+    /// turn, so PC stays <c>ov=0</c> with the seeded counter.
     /// </summary>
     [Fact]
     public void AlphaDepthTranscript_TerraceEdge_MatchesRetailFrame2()
@@ -354,14 +358,19 @@ public sealed partial class WalkTraceConformanceTests
     /// S4-c1 fix round 2, R2-4: the sixth pose (#464 artifact pose, root
     /// <c>0xF4180114</c>, seven outside views). Sixteen building punches —
     /// all <c>mode=1</c>, all <c>counterBefore=0008</c> (far-Z punches never
-    /// touch the counter) — of which twelve are guard-rejected on local
-    /// <c>y=12.000</c> (WalkVisibilityMath's ±12 boundary guard, S4-c1 F1),
-    /// leaving four admitted; then eight true-depth exit seals with
+    /// touch the counter) — then eight true-depth exit seals with
     /// <c>counterBefore</c> running <c>0..7</c> against the root's own
-    /// <c>ov=7</c> read-then-zero, settling this fixture's own steady-state
-    /// counter at <c>8</c> — the same value <see cref="WalkAlphaDepthTrace.LoadInitialCounter"/>
-    /// reads back out of this fixture's own preamble, confirming the seed
-    /// and the replay agree on the same persistent value.
+    /// <c>ov=7</c> read-then-zero. This gate compares only the
+    /// <c>(mode, counterBefore)</c> and <c>(ov, counter, forceClear)</c>
+    /// tuples: it does NOT observe which of the sixteen punches the ±12
+    /// boundary guard rejects (the sink records every attempt before the
+    /// guard runs; the <c>pmverts</c> capture and the DAT, not this Fact,
+    /// show twelve of them lying on local <c>y=12.000</c> — S4 packet §9).
+    /// The seed <c>8</c> is the first PM/PC sample
+    /// <see cref="WalkAlphaDepthTrace.LoadInitialCounter"/> finds in the
+    /// file, which for this fixture lies inside frame 1 (this capture has no
+    /// pre-"F 1" preamble sample); with the priming pass gone, frame 2's PC
+    /// <c>counter</c> equals that seed by construction in steady state.
     /// </summary>
     [Fact]
     public void AlphaDepthTranscript_CathedralStairArch_MatchesRetailFrame2()