Retain PR203 source proof, certificate reconciliation and credit (no new κ)

[steph4n-gh]

Summary

Research consolidation; no new κ.

• Authenticates all 29 paths (195,812 bytes) in huxint's #203 (d2873e20750d0949d2ad75cbc4cd82b5b1a74e84, research/percent-exponent-architecture) offline via raw Git commit, tree, and blob identities in research/pr203-retention-receipt/SOURCE-PROOF.json (60 Git objects: 6 commits, 21 trees, 33 blobs).
• Authenticates the pinned upstream inputs from Rohan Arun's (@rohanarun) #187 (201737a1ec4f936e166e2481fb9e88104cb2ccc7:research/source-assisted-bit-bootstrap/emitted-physical.json, retained via archive/pr-reconciliation100-20261013) and Avi Eisenberg's (@ikeboy) #193 (187e1010ac8b259af8e9b5166f68b64bc27b4b47:research/source-assisted-v4/certificate.json, scripts/experiments/balanced_fano.py, and scripts/experiments/fano_completion.py, retained via archive/pr-reconciliation-next100-20261013).
• Recomputes and reconciles all four exact screens in PR203 against the saved certificates and the Round-9 replay log (research/community-round9-audit/pr203-replay.log.txt):
  1. ceiling-certificate.json: complex-pr193 and bit-pr187 fixed-itinerary and free-center ceilings (a < 0.008868 and a < 0.008898), all 182 finite odd-cube cases (133 positive-deficit cases, scan SHA256 cebf8348bf5dc03f528e809648ed3062340b1d7aaf7b47924e157659f0e374ac, a < 0.005164 for destructive-parity and a < 0.008094 for arbitrary local chains), p >= 30 tail bound, Walsh projector and $K^2 = \frac{1}{4} I$ checks (d in (3, 5, 7)), 5 copied $t$-star binary span checks, and (m=3, W=30) vs (m=3, W=31) one-deficit bit saving bounds.
  2. center-basis-certificate.json: rational Gram invertibility and sharp $\mathbb{F}_2$ support rank $h - 2$ (loss floor $h(h-2)$) for p in (5, 6, 11, 13, 16), p = 4 excluded-domain control (4 < 6), and all $3^{10} - 1 = 59,048$ nonzero ternary coefficient vectors at p = 5 (support rank $\ge 8$).
  3. triangular_obstruction.py: exact common-triangular-flag basis-conjugation and commutativity rejection controls (rank_lower_bound = W * m).
  4. enumeration-controls.json, fano-all-permutations.json, fano-reuse-certificate.json, and discovery/receipts/vector-k*.json: 8 direct full-word brute-force controls, all 87 chronological balanced-Fano stub-reuse topologies, the 16 obstruction-preserving cases matching the 16 PASS k/16 lines in pr203-replay.log.txt, and the 5 block-vector solver receipts ($k = 2, 3, 4$).
• Restores missing PR203 credit in CONTRIBUTORS.md without changing ledger dispositions or claiming a new construction or fresh C++/solver replay.

Task & Coordination

• Assigned task: overlap-receipt (consolidation track, seed 563d7e55de5891491987f38cb6b1508a, scope key 757723e3fd38f20fa774, priority revision 5).
• Base commit: 534de584c7c9b7abd621a0d8cec959d021e5b510.
• Coordination claim: Handoffs and requests for independent checks #1166 (comment)

Verification

python3.12 scripts/research_package.py --package research/pr203-retention-receipt

• Runs verify.py and test_verify.py (21 tests: 1 positive case and 20 coverage, corruption, provenance, ceiling, center-basis, and Fano screen rejection controls) under -I -O -B.
• Verified in an exported directory with no .git and PATH=/nonexistent under Python 3.12.9 and 3.14.4.

Attribution

Credits huxint (PR #203), Rohan Arun (@rohanarun, PR #187), Avi Eisenberg (@ikeboy, PR #193), icekylinx, eumemic (PR #288), an664, DaysSky (PR #192), romainhedouin (PR #183), maxime-fleury (PR #204), and Douglas Colkitt. Original UPSTREAM-NOTICE preserved verbatim. Prepared by steph4n (@steph4n-gh) with substantial Google Gemini assistance. Apache-2.0.