[kumagi]

Change and mathematical scope

Returning bank addresses after intervening data work does not establish a dirty-bank-independent endpoint. This package proves the exact support condition for two controlled Pauli words surrounding an arbitrary data-only complex matrix U, including different consumer labels. It adds necessity to PR2871 P009's sufficient block-diagonal contract and checks the condition in PR2850's actual Fourier control frame.

The unchanged physical producer admits C/Q/translation interpositions. Physical S/Z interpositions return every bank address but leave bank-dependent data maps. A bank Z inserted before the second producer repairs the Z case; putting it after closure fails. No relabelling-only consumer repairs the dense S case. These are proposed interpositions, not bugs in either predecessor's existing emitted programs.

Validation

Evidence: written all-parameter proof within the two-word interface; authenticated unchanged-source replay of 18 physical fixtures at f=1,2 (648 audited basis columns, 37,440 coefficients, plus six comparison programs); separate Gaussian-rational checker with 272 Pauli edge blocks, 16 data matrices, full 8×8 physical matrices and four residual identities. Source corruption, address-only/prepared-bank acceptance, omitted commutation phase and misplaced repair controls are rejected.

Validation passed on Linux x86_64, CPython 3.12.13:

python3.12 scripts/research_package.py --package research/phase-bank-intervening-work-audit
python3.12 scripts/research_package.py --package research/phase-bank-intervening-work-audit --check-only
python3 scripts/check_research_commit.py --package research/phase-bank-intervening-work-audit

Both registered commands use -I -O -B and only the standard library. Sole external runtime input is declared and pinned.

Task and overlap

Task uniform-transfer-audit, refined to one restoration implication for a reviewed phase-bank component. TASK.json records the original card, concrete domain and bottleneck link. Scientific base a7fc79c0cd63d6aa48df753069fc49eabda79dc3; rebased onto canonical main ffbd90dd90dbe0055464f3a1411b25a3a4fdc6b8, with guidance, priority revision 5, selected record and source input unchanged. Fresh complete overlap covered 489 open PRs, 2,092 closed/merged PRs and 1,531 Discussion records; it is not a novelty certificate. This differs from the active PR2872 fixed-label schedule/dimension audit.

Credit

Gareth Davidson/bitplane PR2850 a007c32f992195ad372806db38c9359dd49382ad; Dugongue PR2871 894a3edbcd90fc379a277eb07a5ae2b706c7a8e1 P008/P009/P095. Tekalign/tekuuu PR2872 retains the adjacent paired-schedule results. Substantial OpenAI Codex assistance, including separately delegated proof review and checker implementation; no independent human review or new representation-theory priority claim.

Open next questions

Structured next questions: pin one actual caller, express its intervening work in the Fourier control frame and first test one label-changing support edge. For a transported consumer, charge the extra correction at that call site before extending the mechanism; a favorable complete bill is the blocker.

No selected-frontier change, new κ, supplier admission, native cost bound or selected cold replay. Full caller, joint consumer, all-size compiler, hosting, precision and movement obligations remain. The algebraic Z repair is an added instruction with an unproved completed bill. Scoped claim and handoff: #2842 (comment)