Research: exact screen of mixed rank-six child blocks

[1337isnot1337]

Scoped negative result. No new supplier or κ is claimed.

This package tests the exact block-diagonal fusion (2,2,1,1) -> 6:
diag(A2,B2,a,b) diag(C2,D2,c,d) = diag(A2 C2,B2 D2,ac,bd). The direct sum preserves paid rank mass six while replacing two rank-two and two rank-one children with one rank-six child.

The source-pinned search enumerates all consecutive four-event windows in both complete orientations of the selected gauge44 provider. It finds 28 eligible windows per orientation and a maximum of 13 disjoint fusions. The ideal transformed profile is {1:2386,2:392,3:432,4:142,5:82,6:13,8:8} with mass 5,586. Even after granting zero cost to all source, copy, gather, output, frame, hosting, precision and restoration work, its exact ceiling upper endpoint is 1698446697/1000000000000, below the 5% target 3549113631/2000000000000. At least 126 zero-charge fusions would be required; 125 fails and 126 passes the exact target brackets.

This is limited to the pinned (2,2,1,1) child shape and strict contiguous windows. PR #1779 covers homogeneous rank-one batches; PR #1866 uses a different (3,1,1,1)->6 tensor; open PR #1960 optimizes rank-two/rank-three blocks. This package adds the rank-one/rank-two mixed shape and its source census over both full orientations.

Reproduction

python3 -B scripts/research_package.py --package research/complex-rank6-hetero21-seedfb19 --check-only
python3.12 scripts/research_package.py --package research/complex-rank6-hetero21-seedfb19

The package metadata check and direct verifier pass locally under Python 3.14; Python 3.12 is unavailable on this host. The selected-record check also passes. The registered replay uses Python 3.12 in CI.

Assignment seed fb19d5d7333247ee6fd2818087dc4509, priority revision 3, exact selector base f0c1cd4495674afcc81d618774949690b504217d; selected-record SHA-256 10b9e340fd8e354a37a0c4813c90e5de1dd42044b5f6aca4a3fc5a42b685fa32. Source archives and selected certificates are unchanged.

Prepared with substantial OpenAI Codex assistance in scope selection, analysis, verifier, certificate, exact cost calculation and writing. No independent human review is claimed. New material is Apache-2.0.

[Mister-Meeseeks]

Batch review at a5cb695880f76a4e840357bc58ab6161ee2607ab.

Merged to main, preserving the original contributor commits and package attribution.

Reviewed scope: Both selected orientations contain28 strict mixed-rank windows with maximum13 disjoint fusions. Exact invoice relaxation requires126 such groups for a five-percent gain, so this fixed chronological family fails even with zero adapter cost. The block-diagonal identity is not a physical packing certificate.

The pinned scoped command passed; the batch ledger contains its command, input hashes, environment and log. This incorporates the research/tooling contribution and does not change the selected κ or discharge the inherited multiplication-theorem assumptions.

Batch review, attribution and evidence. Thank you for the contribution.