PMAT-3427: PP-QUANT-001 slice M is 0.69.0 must-carry — measured tables, fix-diff Pareto, the queue

[noahgift]

Docs-only plan PR for the PP-QUANT-001 research run. No edits under crates/, contracts/ or .github/workflows/.

Closes #3427

What lands

• docs/specifications/06x-release-schedule.md §1.1a — the must-carry rule (T-0 waits; more than 72 h past the previous tag raises an andon naming the ticket) and the 0.69.0 row: PP-QUANT-001 slice M + PP-TENSOR-001 T1–T2. It stands in for APR-RELEASE-001 §1.5, which is not on main.
• docs/roadmaps/roadmap.yaml — PMAT-3427 plus 13 tickets minted by the run, each as a docs/roadmaps/entries/<ID>.yaml fragment with the aggregate regenerated (the five per-directory retype tickets were closed into the dedicated-cycle epic and are not mirrored); pmat work validate exit 0.
• docs/audits/impl-PMAT-3427-receipt.md — every number with its mark and command: the five-defect Pareto read from fix diffs, Table G (27 files), Table S (43 ggml ids), the two-sided reconciliation against the dispatch-site list in the PP-QUANT-001 spec PR, the 0.69.0 hygiene count, HRQ rows.

Findings

1. The five cited crashes are two classes: 3 MoE placeholder tensors (byte_size: 0, qtype: 0 = F32) reaching dense consumers — every fix a guard, the placeholder still built on main — and 2 incomplete qtype lists.
2. The 0.68-milestone Qwen3.5 refusal reproduces on the published v0.68.0 asset with a 338 MB public model: exit 8 in tensor_byte_size, type ids 16 and 23. That function has 11 arms and lacks Q3_K and Q5_1, which the crate defines.
3. Three ggml tensor-type enums exist (serve 16 ids, core 12, compute 15) against 35 live upstream; 13 live ids are named by none.
4. 0.68.0 was tagged with 2 issues open in its milestone.

Not done

No quorum reviewed this: the harness rail refuses kind=docs (owned refusal, exit 3), and the brief's design reference was not on disk. Verdict in the receipt is STOPPED(rail): kind-gate exit 3 and model-gate exit 2 on this ticket, verbatim in the receipt. This PR stays draft until a quorum has reviewed the design positions.

Verification

cargo fmt --all -- --check exit 0 · cargo test -p aprender-contracts --lib 1538 passed, 0 failed · cargo deny check advisories ok · pre-commit gates passed.

ont-delta: none — schedule text, roadmap rows and an audit receipt; no ontology row changes.

no-close: #3418 #3421 #3420 #3091 #3428 #3429 #3430 #3431 #3432 #3433 #3434 #3445 #3446 #3204 #2753 #3422 #3423 — cited as plan, evidence or queue; this PR implements none of them.

🤖 Generated with Claude Code

[github-actions]

§13.11 rung 1 — quorum shadow verdict

S13-SHADOW pr=3447 head=d76b4a417e3c2ba3d5edfb23c841a8ad5372f62a verdict=REFUSE class=Q1 arm_rc=1

Shadow mode: this records a verdict and merges nothing. A refusal
to arm is not a block (§13 adds zero rows to §7) — the pull request is
exactly as green as it was.

[noahgift]

stale — APR-RELEASE-001 §6, applied mechanically 2026-09-20.

No activity since 2026-09-17 13:41Z; 2 release trains have been cut since (v0.68.2, v0.68.1). Age is the only input to this sweep — it is not a judgement on the work, and a train label such as pp-066/carry-* is not an exemption.

What happens next: if one more train passes while this is still labelled stale, §6 closes it. The branch is kept — closing a PR here discards a review thread, never commits.

To clear it: push, rebase, or say on the thread what it is waiting for. Any of the three removes the label at the next sweep. If it is blocked on something external, name that here — a blocker with an owner is not sprawl, and it stops the clock.

[noahgift]

Triage disposition (2026-09-20 sweep of DIRTY PRs): needs-an-owner, stays draft on 0.70 — and the slice-M half of its conflict is decided.

This PR is DIRTY on docs/roadmaps/roadmap.yaml (a content conflict on the PMAT-3430 and PMAT-3433 entries — both sides edited the same rows, so the by-id resolver correctly refuses it) and on docs/specifications/06x-release-schedule.md. Its only red check is present, the review-backlog red; CI is otherwise not the blocker.

On the PMAT-3430 row: main's version wins. This PR's edit is from the plan era (09-17). Since then the operator ruled on #3430 (2026-09-20, clauses 1–6 + amendment 2'/2a/2b), #3581 landed the leaf and the upstream-extracted TRAITS[43], and #3621 collapses the three enums into it and closes the ticket. The row here describes a ticket state that no longer exists; a rebase should take main's row as-is for PMAT-3430. PMAT-3433 I have not judged.

On the 06x §1.1a must-carry text: not superseded — grep -i must-carry on main's 06x-release-schedule.md returns nothing — but it is spec prose, and the cut gate is now carried by #3617 rather than by this paragraph. That half is for its owner on 0.70; it is not a cut input.

Not superseded, not stale, not mine to resolve. Refs #3430, #3581, #3621.

[noahgift]

closed by 0.70 batch triage: no quorum receipt and no live owner; branch kept, reopen with an owner

Authority: per operator order (quoted) "yes, clearn up. no sovereign repo can have triage, ticket, branche issues and arbiter shoudl be doing this! tell that agent to get to work" + cop (aprender-cf) ruling on the close criteria (a draft, or >7d old with no live owner, and no quorum receipt). The operator did not select this PR individually.