[aliromman]

This scoped theorem excludes any star caller whose thin connectors have identical columns on three leaves, or identical rows on three leaves. It holds for every finite port dimension, every full-rank orthogonal stage split, every call order, and any number or rank of thin connectors in characteristic two. Complete kappa is unchanged.

Proof, exact evidence, full escape word, source pins.

The proof replaces an order census by a prefix identity. The forbidden chronology would give the same determinant both (1+x)^(3d) and (1+x)^d. Thus adding more pooled queries or more stage splits cannot repair the contract. The resulting signature and rank cut identifies a concrete remaining target: 31 leaves with five rank-two connectors. The corresponding one-port budget is excluded; coupling pairs of ports passes this necessary cut. Its optimistic pinned P104 price is positive at saving .0035, but no such long word or native supplier is supplied.

A different, complete eight-field binary word does escape the obstruction. It has six full calls and eight thin queries. Every polynomial coefficient passes, but its fixed P104 linear margin is -336, and a separated-rank argument rejects every concave moment. A saved counterexample passes the pure stage word and first mixed coefficients but fails higher coefficients. These distinguish a real operator from an insufficient screen.

The separate checker uses no solver or discovery module. Normal and optimized registered checks, five corrupted-package controls, committed-byte checks, and read-only Git-free replay pass. It checks the full escape word, fifteen small paired-order identities, coupled chronology, prefix determinants, and exact integer-power prices. The archive also preserves changed query frames, coupled five-query searches, and four failed fixed-source growth attempts. Timeouts remain unknown; no search failure is used as a universal proof.

Base and refreshed canonical main: df0137efd35e11b4445734bf612bdd4c7ebc330c, priority seven, selected kappa .00310045. Contribution head: c10f75f09e41803ae8efe478331a7509b7379b2f. The bill is frozen at PR #5599, source 09dd0965711ca7eeb169057a420abab1c9052137. This does not change the frontier.

git fetch https://github.com/CrocSwap/integer-mult-bounds.git 0c27282f56a0e8ec6c1fbad0f2ccca646d0d8442
python3.12 -B scripts/research_package.py --package research/crossed-connector-caller

Expected: PASS_ONE_SIDED_CONNECTOR_OBSTRUCTION and PASS_REGISTERED_RESEARCH_PACKAGE. Python 3.12 and the standard library suffice. The fetch supplies an explicitly pinned open provenance object. A full ancestry audit was not run because that source is not an ancestor of the new main base. Hosted CI and independent scientific review are separate statuses. No characteristic-zero reduction, native frame admission, ownership/tape realization, full compiler/outer transfer, or proof-assistant acceptance is claimed.

Credit Cinna Davis's controller/simplex work, PR #5617, PR #5631, and PR #5649. Retain the P104 helper and contact lineage, Gareth Davidson, David Arias, Jacob Alan Sussman, zxv, Cao Yifan, and all pinned source notices. The vendored discovery alphabet retains Cinna Davis's notice. Prepared for Ali Romman with substantial OpenAI Codex assistance; no subagents. Apache-2.0. Coordination: Discussion #5593.

[aliromman]

CI status at c10f75f09e41803ae8efe478331a7509b7379b2f: run 40269950337 failed in the shared plan job. The two failing subcases are test_current_ci_section_matches_selected_record, fields review and source_manifest. The current-CI section lacks links to docs/research/pr-intake-evidence/pr5599-completion/PROOF-REVIEW.md and certificates/frontier5599-source.json. Package selection and hosted replay were skipped.

The tests, scripts, status documentation and selected record are byte-unchanged from base df0137efd35e11b4445734bf612bdd4c7ebc330c; this PR changes only the new research package. This is not a hosted replay pass. Normal and optimized local checks, five corruption controls, committed-byte export and read-only Git-free replay pass at the posted head. Independent scientific review remains open. Final findings and released scope: canonical update.