[unpaidmoon]

Change and mathematical scope

Partial interface result. No new κ. The selected conditional exponent stays 781451/250000000.

On the fixed port9 inverse column-1 leaf, an eight-transition successor marks the raw three-bit new-low with RIGHT and returns that head. It changes no other tape. The marked word fails the leaf's raw-exit contract, so the commit is a later owner step, not a rewrite of the leaf.

Blanking cache bits 24 through 191 (digits 8 through 63) also fails that contract. The published witness row (2, 3, 2, 1, 1, 1, 0, 0) and the same cache with digit 8 changed from 0 to 4 keep the same owner permutation and the same new-low digit 2, and they differ only in the preserved tail. The leaf reads every cache offset 0 through 192 and writes none of them. In both orientations the generic rotation and descriptor cleanup never name tapes 17 or 18.

This leaf is therefore not a legal last consumer of the 64-digit cache. Low-type, ROW and whole-T consumers remain absent.

Task and overlap

Base commit: 2485805 (upstream CrocSwap main).

Item taken: structured handoff #9 in discussion-comment order on #1166, comment 19137812, task supplier-boundary-audit. Its original next task was one N-dimensional stock/owner/route generator after PR #4957. Canonical #4945 has since delivered the route generator, carry frontend, padding controller, quarter rotation and the port9 leaf (PR #6241). Those leave the new-low commit and cache-retirement question open in research/port9-low-cache-carry. This package answers that remaining piece. It does not rebuild the generator.

Scope status: changed. The catalogue card asks about supplier admission coverage. The delivered result is one exit-contract gate on the actual port9 leaf. A passing leaf is not supplier admission, and this gate does not raise κ. It matters because the selected BIT budget still depends on physical cache and output ownership: this leaf may mark its owned new-low and must still return the cache tail.

Nearest overlap: PR #4957, PR #5332, PR #5849 and PR #6241, all used as pinned inputs or predecessors. No open upstream PR title matches this commit/retirement gate.

Validation

CPython 3.12, standard library only, from the repository root:

python3.12 -B scripts/research_selected_baseline.py
python3.12 -B scripts/research_package.py --package research/port9-low-commit-gate --check-only
python3.12 -B scripts/research_package.py --package research/port9-low-commit-gate

Observed: PASS_RESEARCH_BASELINE_BINDINGS_ONLY, PASS_PACKAGE_METADATA_AND_PINS, and PASS_PORT9_LOW_COMMIT_GATE in both registered normal and optimized commands.

The checker reruns both leaf orientations, the digit-8 mutant, the static component tape sets, and both straight-line successors against the predecessor exit test. Finite execution corroborates the static instruction inventory. It is not a formal proof, a whole-T materialization, or make frontier6030-verify (not run).

Credit

Prepared by unpaidmoon with substantial Cursor Grok assistance. Apache-2.0.

AAAmirinu supplied the port9 leaf (PR #6241). JoyboyBrian supplied the low-cache frontend (PR #5332) and the route generator. Ali Romman supplied the complete-owner rotation (PR #5849). David Arias supplied the tape machine. Chris R / Clear1nsight supplied the PR #4957 backend-API handoff. No predecessor bytes or notices are changed.

Open next questions

1. Hand the marked 3-bit word to one real low-type consumer and charge that read. The marker alone is not a consumer.
2. Name the cache's last consumer, then retire bits 24 through 191 only after that read. The missing whole-T layout and ROW installer block any earlier erasure.