diff --git a/crates/sylpheed-formats/examples/coverage_audit.rs b/crates/sylpheed-formats/examples/coverage_audit.rs
new file mode 100644
index 0000000..e49512f
--- /dev/null
+++ b/crates/sylpheed-formats/examples/coverage_audit.rs
@@ -0,0 +1,48 @@
+//! Do real index buffers address their whole vertex pool?
+//!
+//! `validate_block` rejects a block whose indices reach fewer than `vtx_count−4`
+//! vertices ("buffer not covered"). That gate is the furthest-reached rejection
+//! for a handful of resources that never decode — so the question is whether it
+//! is well founded. This measures the slack on every block that DOES decode: if
+//! real geometry always covers its pool, under-coverage is good evidence of a
+//! wrong candidate and the gate stands.
+use sylpheed_formats::mesh::Xbg7Model;
+use std::collections::BTreeMap;
+
+fn main() {
+    let dir = std::env::args().nth(1).expect("resource3d dir");
+    let mut files: Vec<_> = std::fs::read_dir(&dir)
+        .unwrap()
+        .flatten()
+        .map(|e| e.path())
+        .filter(|p| p.extension().and_then(|s| s.to_str()) == Some("xpr"))
+        .collect();
+    files.sort();
+
+    let mut hist: BTreeMap<i64, usize> = BTreeMap::new();
+    let mut worst: Vec<(i64, String)> = Vec::new();
+    for f in &files {
+        let Ok(bytes) = std::fs::read(f) else { continue };
+        for m in Xbg7Model::anchor_models_cancellable(&bytes, 0.0, &|| false) {
+            for sub in &m.meshes {
+                if sub.positions.is_empty() || sub.indices.is_empty() {
+                    continue;
+                }
+                let max_idx = *sub.indices.iter().max().unwrap() as i64;
+                let slack = sub.positions.len() as i64 - 1 - max_idx;
+                *hist.entry(slack.min(20)).or_default() += 1;
+                if slack > 4 {
+                    worst.push((slack, format!("{} in {}", m.name, f.file_name().unwrap().to_string_lossy())));
+                }
+            }
+        }
+    }
+    println!("unreferenced tail vertices (vtx_count − 1 − max index), over decoded sub-meshes:");
+    for (slack, n) in &hist {
+        println!("  {:>3}{} : {n}", slack, if *slack == 20 { "+" } else { " " });
+    }
+    worst.sort_by_key(|(s, _)| std::cmp::Reverse(*s));
+    for (s, w) in worst.iter().take(5) {
+        println!("  largest slack {s}: {w}");
+    }
+}
diff --git a/crates/sylpheed-formats/src/mesh.rs b/crates/sylpheed-formats/src/mesh.rs
index 563deec..92a952c 100644
--- a/crates/sylpheed-formats/src/mesh.rs
+++ b/crates/sylpheed-formats/src/mesh.rs
@@ -1391,7 +1391,17 @@ fn anchor_grouped_meshes(
                 if !ok && i > kmax {
                     break; // chain diverged — emit the validated prefix, no garbage
                 }
-                meshes.push(read_pool_mesh(bytes, ib, vb, ic, vc, decl));
+                // Sub-meshes BEFORE the pivot are emitted even when they fail the
+                // quality gates (a tiny flat lead part is legitimately poor), but
+                // an index that addresses past its own vertex buffer is not a
+                // quality question — it is unusable. Measured 2026-08-12: 18
+                // sub-meshes disc-wide carried indices up to 364 vertices past
+                // the end (`coverage_audit`), which any renderer would fault on.
+                let in_range =
+                    (0..ic).all(|k| (be16(bytes, ib + k * 2) as usize) < vc);
+                if in_range {
+                    meshes.push(read_pool_mesh(bytes, ib, vb, ic, vc, decl));
+                }
                 vb += vc * stride;
             }
             return meshes;
diff --git a/docs/re/structures/xbg7-mesh.md b/docs/re/structures/xbg7-mesh.md
index 652a72d..8b0db1b 100644
--- a/docs/re/structures/xbg7-mesh.md
+++ b/docs/re/structures/xbg7-mesh.md
@@ -965,6 +965,36 @@ container holds two direct copies **and** two mirrored ones
 (`0x33b7754`, `0x342e284`); our twins take the two direct copies, which is
 self-consistent but unverified — `n206` appears in no captured stage.
 
+### ✅ Fixed: grouped pools emitted sub-meshes with out-of-range indices
+
+Chasing whether the "buffer not covered" gate is well founded turned up a real
+defect instead. `examples/coverage_audit.rs` measures, for every decoded
+sub-mesh, how many tail vertices its indices never reference:
+
+| unreferenced tail vertices | sub-meshes |
+|---|---|
+| **0** (indices reach the last vertex exactly) | **8 586** |
+| 1–3 (inside the gate's ±4 tolerance) | 48 |
+| 9, 80 (`f102_break.dat`, `f104_break.dat` in `ptc_pack.xpr`) | 2 |
+| **negative — indices point PAST the vertex buffer** | **18** |
+
+The first row answers the original question: real geometry covers its pool
+**exactly**, so under-coverage is good evidence of a wrong candidate and the gate
+stands as written.
+
+The last row is the defect. `anchor_grouped_meshes` reads sub-meshes *before* the
+pivot unconditionally — deliberately, since a tiny flat lead part legitimately
+fails the quality gates — but that also skipped the **index-range** check, which
+is not a quality question. Eighteen sub-meshes disc-wide were emitted with
+indices reaching up to **364 vertices past the end** of their own buffer, which
+any renderer would fault on or draw as garbage. Pre-pivot sub-meshes are now
+required to be in range (quality gates still relaxed); out-of-range ones are
+dropped and the pool's remaining parts are kept.
+
+Everything else holds: 6 069/6 294 decoded, cross-container inconsistency 56, the
+capture truth table still 46/46 claimed with 0 unclaimed, suite green. The vertex
+total falls by 1 546 — exactly the garbage that is no longer emitted.
+
 ### Coverage has a denominator now, and the misses have a cause breakdown
 
 Coverage has been quoted as "resources decoded" with no total. `examples/undecoded.rs`