Independently audit QQ16 intervention obstruction and emitted route bindings

[EdDiboi]

Change and mathematical scope

Independent audit of PR #2306's deterministic selected-only QQ16 intervention obstruction. A different eight-variable integer difference module corroborates its 19-event argument, and a separate checker binds 14,784 emitted alignment references to their actual stage and port (2,752 in the retained nine cases).

The same nonzero initial-A difference survives every permitted deterministic selected-only update, including affine, nonlinear and adaptive updates: identical permitted observations and auxiliary states yield identical updates. The proof is an indistinguishability argument, not a finite test of nonlinear controllers.

No new witness, controller, supplier, paid saving or κ is claimed. Logical/native-frame and bridge conventions are inherited; intermediate ordinary reads, pre-F0 retention, and former-Y endpoint reads are outside this obstruction. No selected-frontier files change.

Task and overlap

• Base main: 0fdd7d82de03e43c9e5a1000dfbcd380d1b687d2; priority revision 4.
• Task: joint-query-controller, scope key 0e34adbc1dd119161c3a; scope deliberately changed to independent verification because live construction follow-ups were claimed.
• Audited predecessor: research: rule out selected-only QQ16 boundary repairs #2306 at 4d6e08099c11ff06a8585d04ed59453dd5a62d7d.
• Claim: CLAIM: QQ16 F0-to-B1 selected-state shear with existing query width (seed ccce426f) #2205 (comment)
• Distinct from Research: no single coordinate tap restores the QQ16 selected inputs #2326's coordinate-tap kernel and Research: the QQ16 shear prefix needs exactly sixteen native lines #2328's native prefix pricing. Source pins and overlap limits are in TASK.json and SOURCES.json.

Added value is the reduced transfer certificate and semantic route association, beyond unordered registry membership. Existing source hashes already reject changed inputs; this does not report an incorrect frozen route or a security defect.

Validation

CPython 3.12.14, standard library:

python3 -I -B scripts/research_package.py --package research/qq16-intervention-independent-audit

Registered normal and optimized replays pass; metadata/pins and git diff --check pass. RESULT.json retains the full 8×8 tail matrix and event states. Six controls reject wrong signs/pair/bridge/terminal action and two registry-preserving wrong alignment references; identical-matrix aliases pass.

The published seven-file tree was fetched and compared byte-for-byte with the locally tested tree. Full supplier cold replay, native scalar/private-copy execution, hosted CI, Lean, all-size and multiplication-theorem verification were not run or claimed.

Credit

Adam Gall (#2093): three-port witness; Ali Romman (#1969): QQ16 maps/routes/schedule; Cinna Davis (#2216): older-witness shear; @CubeJerry (#2306): deterministic intervention obstruction. @commdata2338's recovery work and the unchanged supplier/compiler notices retain credit.

Prepared by @EdDiboi with substantial OpenAI Codex assistance in implementation, proof review, verification and writing, including separate agent review. This is independent agent verification, not independent human review or maintainer acceptance. New work is Apache-2.0.

Open next questions

See the package README: directly adopt the two emitted-reference mutation fixtures in the predecessor checker, or investigate a legally priced ordinary observation/pre-F0 retention against the complete kernel. Native access, full injectivity, restoration and the complete bill remain open.

[Copilot]

Copilot was unable to review this pull request because the user who requested the review has reached their quota limit.