diff --git a/.agents/issue-index.md b/.agents/issue-index.md
index 519a50241..d0a36cb88 100644
--- a/.agents/issue-index.md
+++ b/.agents/issue-index.md
@@ -480,3 +480,5 @@ rather than merged. `scripts/check-agent-record.py` gates both.
 | [#1435](https://github.com/mudler/vllm.cpp/issues/1435) | `VT-MATMUL-FP8-BLOCK-CUDA` | **A CUDA build without CUTLASS headers segfaults on the block-wise FP8 path instead of refusing by name, and configure reports the feature ENABLED anyway.** Measured 2026-08-20 on `dgx:gpu0` (GB10, cc 12.1) in an `rc` lease at `63d87805c`. `src/vt/cuda/cuda_matmul_fp8_block_cutlass.cu` is the only TU that registers `kMatmulFp8BlockScaled` for `kCUDA` and it reaches `target_sources(vllm PRIVATE ... ${_FP8_CUTLASS_SOURCES})` only through TWO nested conditions in `CMakeLists.txt` -- the outer `if(VLLM_CPP_CUTLASS)` (one occurrence in the file) and, inside it, the `if(VT_CUTLASS_FP8_ARCHS)` that SETS `_FP8_CUTLASS_SOURCES` -- so with no CUTLASS headers the op is unregistered, a device tensor reaches the portable HOST kernel, and the process SIGSEGVs while the reference tier prints `correct but slow`. Anchored by SYMBOL rather than by line on purpose, and the reason is measured on this very row: the guard sat at `CMakeLists.txt:1695@63d87805c` -- the only line number in this row, sha-qualified so it stays resolvable -- and by the time this row was written that number was already stale, then rotted AGAIN by four lines when `origin/main` was merged into the row's own branch mid-review. Two rots inside one pull request. A bare head-relative number here would be permanent, because this index is append-only and no checker scans it, and it would resolve to plausible unrelated code rather than to nothing: at `63d87805c` that line was the guard, and two trees later it was an NCCL include directory. Resolve the anchors above by grepping the symbols; do not trust any number quoted for them. Reachable on a default build: `VLLM_CPP_CUTLASS_FETCH` defaults OFF (the single `option(VLLM_CPP_CUTLASS_FETCH ...)` line, `OFF` at its default argument) and CUTLASS is not a submodule. This contradicts the design note in `cuda_matmul_fp8_block_cutlass.cu`, which states an unsupported build `keeps refusing by name -- which is the honest answer and not the #960/#844 fall-through`: the refusal IS real on the model path (`RefuseUnrunnableQwen3_5DenseFp8Block`, plus the two `VT_CHECK`s) but a caller entering through the op registry, which is what the ported upstream test does, bypasses all three. Second, smaller defect in the same run: `vt_cuda_report_feature` printed `CUDA feature cutlass-fp8: ENABLED for [121a]` for a build whose `build.ninja` contains ZERO references to either FP8 CUTLASS TU, because the report runs on the arch intersection alone, before CUTLASS-header detection. `cmake/CudaArchFeatures.cmake` opens by naming this exact class as what the feature table exists to prevent. NOT fixed in flow: both are code changes on a CUDA path needing their own row, spec and hardware re-gate. Owned by row `VT-MATMUL-FP8-BLOCK-CUDA`, under `## Owed` in [`vt-matmul-fp8-block-cuda.md`](specs/vt-matmul-fp8-block-cuda.md) | bug |
 | [#1437](https://github.com/mudler/vllm.cpp/issues/1437) | `VT-MATMUL-FP8-BLOCK-CUDA` | **The mainloop-scaled CUTLASS block-wise FP8 GEMM throws `cutlass Invalid status` on upstream's own ported case, on the arch it targets.** FIRST on-hardware execution of #1189 M5 (`489a9a4c0`), measured 2026-08-20 on `dgx:gpu0` (GB10, driver 580.173.02, compute capability 12.1) in an `rc` lease at `63d87805c`, CUDA 13.0 Release, `-DVLLM_CPP_CUDA_ARCHITECTURES=121a -DVLLM_CPP_CUTLASS_FETCH=ON`. G2 (upstream's `test_w8a8_block_fp8_cutlass_matmul` ported whole: M=32, N=576, K=7168, block [128,128], bf16 out) and G7 both THROW from `gemm_op.can_implement(args)`, so CUTLASS rejects the configuration before any launch -- not a numerical disagreement and not a launch failure. G6, G8 and G9 PASS. That this was an EXECUTION and not a skip rests on the throw TEXT, not on a count: `vt cuda: matmul_fp8_block_scaled: cutlass Invalid status` is assembled by the `VT_CUTLASS_CHECK` macro in `src/vt/cuda/cuda_matmul_fp8_block_cutlass.cu`, and the message prefix `matmul_fp8_block_scaled: cutlass ` occurs in NO other translation unit (the per-tensor sibling `cuda_matmul_fp8_cutlass.cu` defines a same-named macro but writes a different message), in a TU compiled only when CUTLASS is found, so no host fallback can produce that string; the run also logged ZERO `[vt reference-tier]` lines, and artifact provenance was confirmed by `cuobjdump --list-elf` reporting `cuda_matmul_fp8_block_cutlass.cu.1.sm_121a.cubin`. Suite: 5 cases, 3 passed, 2 failed, **34 assertions, 0 failed** -- recorded, but attributed rather than leaned on, because **27** of the 34 are G6, which has no `HasCuda()` guard and prints identically on a GPU-less box (8 grid entries x 3 `CHECK` + 3 tile-config `CHECK`s); the device-only remainder is G8's 2 and G9's 5, and a pure skip of this file prints 27. NO shape has had its output compared with the CPU reference: G2 and G7 are the only two cases that make that comparison and BOTH threw before their first assertion -- G7 aborted at `Grid()[0]`, the same M=32 N=576 K=7168 case, so its other seven shapes, including the second ragged-N entry M=8 N=576 K=1024, were never attempted -- and G8 compares a device f32 out against a device bf16 out, which is the kernel against ITSELF. Hypothesis recorded in the issue and explicitly NOT isolated: N=576 is 4*128+64, a ragged final scale block, which the test header records as upstream's reason for choosing it; G8 passing makes the failure shape-dependent rather than universal. This moves the row from `unmeasured` to `measured and failing`, a worse position than the spec previously described. NO speed claim: the lease took no clock control and recorded no contention. Owned by row `VT-MATMUL-FP8-BLOCK-CUDA`, under `## Owed` in [`vt-matmul-fp8-block-cuda.md`](specs/vt-matmul-fp8-block-cuda.md) | measurement |
 | [#1447](https://github.com/mudler/vllm.cpp/issues/1447) | `ENG-EXPERT-STREAM` | **`docs/USAGE.md` said EVERY Qwen3.8 decode figure came from the W0e C ABI harness; the 66.7 s/token streaming-off row of 16 August 2026 came from `vllm-server`, the same binary the section tells the reader to run.** Introduced by #1211 and fixed in the same flow: the sentence is scoped to the W0e and W0f runs and the exception is named, in `docs/USAGE.md` and in the spec paragraph that mirrors it. Provenance read at the source, `.agents/specs/expert-streaming.md:837` (server entry point) and `:905`. | record |
+| [#1403](https://github.com/mudler/vllm.cpp/issues/1403) | `ENG-CUDAGRAPH-BREAK` | **`test_qwen3_5_decode_graph_seam` SIGSEGVs — a DUPLICATE of [#1394](https://github.com/mudler/vllm.cpp/issues/1394) that was already fixed when it was filed, so no code change was owed.** Filed against `96ed8346f`, which PREDATES the fix `7dec1d990` (PR #1393, merged 2026-08-20 00:49); `git merge-base --is-ancestor` puts the fix outside the reported tree. Re-measured at `b537a5344`, CPU-only Release x86_64: the test passes 5 runs of 5, `8 cases`/`8 passed`, `assertions: 138`, `Status: SUCCESS!`, exit 0. The cause was re-derived by single-variable mutation rather than inherited from #1394's report, and NAMED by a sanitizer rather than reasoned. Debug + `VLLM_CPP_SANITIZE=address,undefined` (NDEBUG OFF) with BOTH halves of `7dec1d990` reverted to their `96ed8346f` bytes: `AddressSanitizer: SEGV on unknown address 0x5045f5f84900`, `The signal is caused by a READ memory access`, at `src/vt/cpu/cpu_paged_attn.cpp:59` in `KvElem<KvKind::K>` reached from `:224`, on threadpool worker T1, inside case `W6: two spec shapes of EQUAL S and different q get two graphs` (`tests/vllm/models/test_qwen3_5_decode_graph_seam.cpp:800`). **It is NOT a use-after-free and there is no free site**, and `AddressSanitizer can not provide additional info` is the tell: the out-of-bounds block-table read is itself IN-BOUNDS for ASan because it lands inside the neighbouring pooled allocation, so this is not a `heap-buffer-overflow` either — the VALUE read becomes a wild block index that `KvElem` multiplies by the KV block stride and dereferences, which is why the fault address is unmapped rather than a redzone and why the crash is order-dependent instead of deterministic. Both halves are load-bearing: reverting ONLY the `SpecAttnMeta` fixture leaves the kernel bound to refuse by name at `src/vt/cpu/cpu_paged_attn.cpp:152` (exit 1, the case THROWS); reverting the bound as well restores the crash (Release exit 139, 3 runs of 3; ASan exit 1); restoring both gives exit 0, 8/8, zero sanitizer findings. The issue's `5f68e60df` pointer was right about the commit and wrong about the half — `git log -S'SpecAttnMeta'` and `git log -S'W6: two spec shapes of EQUAL S'` each return that commit ALONE, and both hits sit in the TEST fixture, not in the graph eligibility it widened. [#1405](https://github.com/mudler/vllm.cpp/issues/1405)'s truncation reproduced (`assertions: 135` crashed vs `138` complete) and is SHARPER under a threadpool fault: doctest printed the summary TWICE with different totals, `135` then `141`, because the main thread kept running after a worker died. That a red reached `main` unreported is [#1285](https://github.com/mudler/vllm.cpp/issues/1285)/[#1376](https://github.com/mudler/vllm.cpp/issues/1376), not this row | bug |
+| [#1458](https://github.com/mudler/vllm.cpp/issues/1458) | `VT-ACT-ROUND-POLARITY` | **`4712dac40` reds FOUR suites on `main` — `test_ltx2_text_encoder`, `test_muse_glimmer_text`, `test_muse_glimmer_text_fallback`, `test_minimax_music3_ar` — by exceeding bf16 error floors none of them had re-derived.** Found while gating [#1403](https://github.com/mudler/vllm.cpp/issues/1403) (PR #1457) and filed in flow, not that row's defect. Deterministic, not a load artifact: first seen under `-j4` and re-measured SERIALLY, 101 s across the four. `test_ltx2_text_encoder.cpp:2407` reads `CHECK( 0.1323 <= 0.109394 )` and `:2409` reads `CHECK( 0.0752773 <= 0.0573374 )`, both over by 20-30%, with the file at `26 passed / 1 failed`, `4118 assertions`, `Status: FAILURE!`. ATTRIBUTED BY MUTATION rather than inferred, both directions, in one build directory changing only `src/vt/cpu/cpu_ops.cpp`, compile rc 0 on every arm and `sha256sum` taken before and after: at `0adeb8b0e` four fail; reverted to `4712dac40^` four pass and `test_ltx2_text_encoder` is 27/27 `SUCCESS!`; restored to a re-matched `sha256` four fail again. They also passed at `b537a5344`, three commits earlier, in a full 567-test `ctest` run at 3-6 s each. `4712dac40` is +42/-3 in `src/vt/cpu/cpu_ops.cpp` plus a new `tests/vt/test_ops_activation.cpp`, and touches none of the four. NOT FIXED IN FLOW, deliberately: the question is a numerics decision, not a defect with an obvious repair — either the floors were calibrated against the rounding polarity that commit corrected and need re-deriving against the oracle, or the narrowing is wider than upstream's — and editing the four floors to pick the first branch is the scope-widening `AGENTS.md` prohibits | bug |
diff --git a/.agents/specs/eng-cudagraph-break.md b/.agents/specs/eng-cudagraph-break.md
index a4df96ca4..ec5f732d8 100644
--- a/.agents/specs/eng-cudagraph-break.md
+++ b/.agents/specs/eng-cudagraph-break.md
@@ -2590,3 +2590,53 @@ now — and this change neither created nor widened it. Owner: row
 **`ENG-CUDAGRAPH-BREAK`**, the stage that gets a `dgx` window WITH the
 Qwen3-0.6B/4B checkpoints, which is the same window the decline and the depth-2
 battery already owe runs to.
+
+**A duplicate of the block-table defect, and why the sanitizer describes it
+badly** ([#1403](https://github.com/mudler/vllm.cpp/issues/1403), a re-report of
+[#1394](https://github.com/mudler/vllm.cpp/issues/1394)). `#1403` was filed
+against `96ed8346f` while the fix `7dec1d990` was in flight, so it names a red
+that `main` no longer carried by the time anybody read it. No code was owed, and
+the useful part is what the re-derivation measured rather than the conclusion it
+reached.
+
+The obvious reading of the first report is wrong in a specific way worth keeping.
+`5f68e60df` is the right commit to start from — `git log -S'SpecAttnMeta'` and
+`git log -S'W6: two spec shapes of EQUAL S'` each return it alone — but the
+defect is in the TEST FIXTURE that commit added, not in the graph eligibility it
+widened. A reader who starts at the eligibility predicate is reading production
+code that was correct, and the widening is only what first drove a shape through
+the fixture that the fixture could not describe.
+
+**Both halves of `7dec1d990` are load-bearing, which its own gate could not show.**
+That commit repairs `SpecAttnMeta` and adds the kernel bound, and after the repair
+its gate no longer reaches the bound — the point
+[#1407](https://github.com/mudler/vllm.cpp/issues/1407) makes when it moves the
+refusal's gate to `tests/test_ops_paged_attn`. Reverting the two halves
+separately, in this binary, closes that: with only the fixture reverted the bound
+refuses by name at `src/vt/cpu/cpu_paged_attn.cpp:152` and the case THROWS
+(exit 1); with the bound reverted as well the crash returns (Release exit 139,
+three runs of three); with both restored the file is exit 0, 8/8, 138 assertions.
+
+**The sanitizer calls this a SEGV, not a `heap-buffer-overflow`, and that is not
+a limitation of the tool.** Debug plus `VLLM_CPP_SANITIZE=address,undefined`
+(NDEBUG OFF) on the fully reverted tree reports
+`AddressSanitizer: SEGV on unknown address 0x5045f5f84900`,
+`The signal is caused by a READ memory access`, at
+`src/vt/cpu/cpu_paged_attn.cpp:59` in `KvElem<KvKind::K>` reached from `:224`, on
+threadpool worker T1. There is NO free site to report, because nothing is freed:
+the out-of-bounds block-table read is IN-BOUNDS for ASan, landing inside the
+neighbouring pooled allocation, and it is the VALUE read that becomes a wild
+block index which `KvElem` multiplies by the KV block stride and dereferences.
+The consequence for the next reader is that **a clean ASan report at the table
+read is not evidence the read is in bounds** for this defect class. The pool
+places the tables, so ASan sees one live region; only the bound at `:152` knows
+where the table's own last column is. It also explains the order dependence
+[#1407](https://github.com/mudler/vllm.cpp/issues/1407) measured: whether the
+neighbouring bytes decode to a mapped address is the allocator's business.
+
+The truncation in [#1405](https://github.com/mudler/vllm.cpp/issues/1405) is
+worse than a short count when the fault is on a worker thread. Doctest printed
+its summary TWICE with different totals, `135` then `141`, because the main
+thread kept running after T1 died. A gate that pins a total therefore sees count
+drift, and a gate that greps for one summary line can read whichever of the two
+it reaches first. The exit code remains the authority.