fix(mesh): distinct assignment now covers grouped pools too

anchor_grouped_meshes takes the same taken set; a colliding grouped model is
re-placed whole past everything claimed, or keeps what it had. Runtime oracle
unchanged (45/46 exact, 0 unclaimed), coverage unchanged (6069), cross-container
inconsistency 62 -> 56, suite green.

It does not clear the n206 twin collapse: no alternative pool validates for the
loser, so that pair is a validator case (like eng_02_l before the cap move), not
a selection one. Also fixes a stale '0.28 (default)' label in edge_cap_sweep.

Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>

crates/sylpheed-formats/examples/edge_cap_sweep.rs

@@ -75,6 +75,6 @@ fn main() {
     }
     println!(
         "cap={}  models={models}  verts={verts}  shared={shared}  inconsistent={inconsistent}",
-        std::env::var("XBG7_EDGE_CAP").unwrap_or_else(|_| "0.28 (default)".into())
+        std::env::var("XBG7_EDGE_CAP").unwrap_or_else(|_| "0.42 (default)".into())
     );
 }

crates/sylpheed-formats/src/mesh.rs

@@ -589,7 +589,8 @@ impl Xbg7Model {
             } else {
                 // Several sub-meshes sharing grouped index/vertex pools → the
                 // deterministic grouped-pool decode (hero ships et al.).
-                let grouped = anchor_grouped_meshes(bytes, data_base, starts, &r.markers, &r.decl);
+                let grouped =
+                    anchor_grouped_meshes(bytes, data_base, starts, &r.markers, &r.decl, &empty_taken);
                 if !grouped.is_empty() {
                     grouped
                 } else {
@@ -652,10 +653,10 @@ impl Xbg7Model {
         let mut models: Vec<Xbg7Model> = Vec::with_capacity(decoded.len());
         for (i, mut m) in decoded {
             let r = &resources[i];
-            if m.meshes.len() == 1 && r.markers.len() == 1 {
+            let starts = &starts_by_stride[&r.decl.stride];
             if let Some(vb) = m.meshes[0].vbuf_offset {
                 if taken.contains(&vb) {
-                        let starts = &starts_by_stride[&r.decl.stride];
+                    if m.meshes.len() == 1 && r.markers.len() == 1 {
                         let (vtx_count, index_count) = r.markers[0];
                         if let Some(alt) = anchor_pool_mesh(
                             bytes,
@@ -668,10 +669,21 @@ impl Xbg7Model {
                         ) {
                             m.meshes[0] = alt;
                         }
-                        // No free candidate → keep the collided decode rather
+                    } else {
-                        // than drop the resource; coverage never regresses.
+                        // Grouped pool: re-place the WHOLE pool past everything
+                        // claimed, or keep what we had.
+                        let alt = anchor_grouped_meshes(
+                            bytes, data_base, starts, &r.markers, &r.decl, &taken,
+                        );
+                        if !alt.is_empty() {
+                            m.meshes = alt;
                         }
-                    if let Some(vb2) = m.meshes[0].vbuf_offset {
+                    }
+                    // No free candidate → keep the collided decode rather than
+                    // drop the resource; coverage never regresses.
+                }
+                for sub in &m.meshes {
+                    if let Some(vb2) = sub.vbuf_offset {
                         taken.insert(vb2);
                     }
                 }
@@ -1147,6 +1159,7 @@ fn anchor_grouped_meshes(
     starts: &[usize],
     markers: &[(usize, usize)], // (vtx_count, idx_count) in descriptor/file order
     decl: &VertexDecl,
+    taken: &std::collections::HashSet<usize>,
 ) -> Vec<GameMesh> {
     let n = markers.len();
     if n == 0 {
@@ -1182,6 +1195,11 @@ fn anchor_grouped_meshes(
     let (vck, ick) = markers[kmax];
 
     for &vb0 in starts {
+        // Distinct assignment: a pool another resource already claimed is not a
+        // candidate (see the collision resolution in `anchor_models_filtered`).
+        if taken.contains(&vb0) {
+            continue;
+        }
         for pad in 0..=3usize {
             if vb0 < span + pad {
                 continue;

docs/re/structures/xbg7-mesh.md

@@ -930,6 +930,29 @@ of one buffer, a resource landing on a **mirrored** copy would be a real defect
 and would look identical to a correct decode in every count-based metric. Only a
 capture (or the twin-pair invariant) can catch it.
 
+### Distinct assignment extended to grouped pools (2026-08-12)
+
+`anchor_grouped_meshes` now takes the same `taken` set: a pool whose start another
+resource already claimed is skipped, and a grouped model that collides is
+**re-placed whole** past everything claimed (or keeps what it had, so coverage
+cannot regress).
+
+| | oracle exact / 46 | unclaimed | resources decoded | shared inconsistent |
+|---|---|---|---|---|
+| single-mesh distinctness only | 45 | 0 | 6 069 | 62 |
+| **+ grouped pools (now)** | 45 | 0 | 6 069 | **56** |
+
+The runtime oracle is unchanged and cross-container inconsistency falls another
+10 %. The suite, including the disc- and ISO-gated tests, stays green.
+
+**It does *not* clear the `n206` collapse**, and that is informative: both twins
+are grouped, the loser is re-placed past the taken pool, and no alternative pool
+**validates** — so it keeps the collided decode. `n206_02` is therefore the same
+class as `e106_eng_02_l` was before the cap moved: the correct block is rejected
+by the validator, not lost to selection. Its correct pool is one of
+`0x342d984` (direct) or `0x33b7754` / `0x342e284` (mirrored); a capture of a
+stage containing `n206` would say which, and is the cheapest way to settle it.
+
 ### The twin invariant, checked disc-wide (2026-08-12)
 
 The capture gave a rule that needs no capture to apply: a `…_01`/`…_02` pair of