[K3] Independently replay the cell-5 xi3 pairing frontier

[scottdhughes]

Summary

Restacked on the current #1152 head. This ready-for-review packet gives an independent, public-source replay of the six residual xi=3 cell-5 pairing representatives.

• Pairings 3/4/5: 20 signed rows, 120 target roots, 296 candidates, 432 source routes, zero terminal solutions, zero witnesses, zero unresolved branches.
• Pairings 7/8/11: 24 signed rows, 140 target roots, 308 candidates, 312 source routes, 40 lifted z/q candidates, zero terminal solutions, zero witnesses, zero unresolved branches.
• Both packets are exact over F_2130706433 and pin public-DAG commit 28b3bc8a plus every compiler/tower/kernel SHA-256.
• Wolfram independently checks a representative nonsquare no-lift and the common quadratic-resultant/sign-free identities.

Relationship to #1152

#1152 subsequently added a larger 23-node PROVED packet claiming the full 210-label [5,8] closure. Its pinned public-DAG commit 3fa298743024 was not fetchable during the 2026-08-09 independent audit: direct fetch returned not our ref, GitHub returned No commit found, the named branch was absent from ls-remote, and the sampled closure nodes were absent from current public master.

This PR therefore remains useful as a reproducible independent cross-check of the xi3 cell-5 subfrontier. It does not replace or validate #1152 endpoint-rootlessness or cell-5-to-cell-8 transport nodes. The audit verdict is RED only for current provenance replayability, with no verdict on the truth of #1152 mathematics.

Verification

• Pairings 3/4/5 certificate: PASS; 10/10 mutations rejected.
• Pairings 7/8/11 certificate: PASS; 12/12 mutations rejected.
• Python compile: PASS.
• Fresh source-bound case-23 replay: COMPLETE; zero witnesses; committed digest match.
• Wolfram symbolic checks: PASS.
• Restacked diff check: PASS.

Status and nonclaims

Ready for independent review, but both local packets remain EXPERIMENTAL_REVIEW_REQUIRED / YELLOW. Ledger movement is zero. This does not close K3, the KoalaBear row, or the global S_sparse gate, and it does not use the falsified FLOOR-v2 random-word first-moment route.