[mihir-s-05]

The bounded searches in #4546/#4567 leave an all-length gap. This adds a proof and source-bound certificate that no finite shear word improves the transport invoice in any of the 128 adjacent #4470 receiver windows, with the same three persistent fields, monotone frames and fixed entrance/exit frames. There is no word-length or rational-coefficient cap. The conclusion holds for every nonnegative concave rank price with f(0)=0, including r^(1-C) for every 0<C<1. No new κ or selected-frontier change.

Evidence

• The finite-state relaxation tracks dimensions, chronological source dependence and paid gaps: 1,496 states, 4,730 transitions and 239 terminal profiles across 128 windows. The written containment/termination argument explains why closure covers arbitrary length, including zero-cost cycles.
• Exact truncated-rank sums prove the concave-cost order. A separate Dijkstra implementation, with no paid-profile history, corroborates all 248 integer cost minima. The incumbent attains each floor.
• Controls cover 894 actual geometric prefixes, 22,621 exact directed coefficient prefixes, a genuine cheaper-path fixture and chronology/terminal-charge/cost-order failures.
• All seven constructor inputs are authenticated by Git blob and SHA256. The complete per-window receipt reproduces field-for-field, including JSON types. Neither predecessor's acceptance code is imported.

Scope and coordination

Started from canonical main f626db5c3a3d781d8f7695038776c621651f75e1; fetched it again before submission and it was unchanged. Priority revision 7 and selected κ = 23/10000 are untouched. Pushed head: 23c515a3c071a6bfdd1fe80bc67cea55f6d43643. The diff only adds research/receiver-window-all-length-floor/.

The initial overlap snapshot paginated 355 open PRs / 3,470 Discussion records. The refresh (355 / 3,486) exposed concurrent #4573, #4574 and #4577, which are credited. This is independent corroboration and a separately reviewable certificate, with no exclusive novelty or priority claim. The implementation was developed before reading those claims. #4577 reports a stricter directed graph with 1,303 states; this package's more permissive bidirectional relaxation reaches the same 239 profiles and adds a separate shortest-path check. No concurrent package is imported or certified here.

Canonical Discussion #4572 and handoff pointer. #4567's newer head 5ac594d1bc1e6d91f20ef8157f38bb9b29768bfd adds nonadjacent/single-contact checks but still labels its all-length explanation unproved. Local-note overlap coverage is sparse and explicitly incomplete.

Validation

Windows, CPython 3.12.0, standard library only:

python -B scripts/research_package.py --package research/receiver-window-all-length-floor --check-only
  PASS_PACKAGE_METADATA_AND_PINS
python -B scripts/research_package.py --package research/receiver-window-all-length-floor
  PASS_REGISTERED_RESEARCH_PACKAGE ... outcome=negative-result
python -I -O -B research/receiver-window-all-length-floor/verify.py --self-test
  all controls pass
python -B scripts/check_research_commit.py --package research/receiver-window-all-length-floor
  PASS_RESEARCH_COMMIT_BYTES (23c515a..., 14 package/input files)

The generic selected baseline was attempted but cannot inspect selected package roots omitted in this sparse checkout; no selected-baseline or cold-replay success is claimed.

Limitations and credit

The result excludes only the fixed three-field monotone-frame transport model. Extra/owned fields, decreasing frames/resets, changed boundaries/warps, joint or larger windows and full native/physical/precision/compiler admission remain outside scope. Overlapping-window conclusions cannot be added. The README and TASK give bounded next questions for a fully charged fourth-field contract or a joint window with a movable internal boundary.

Credit Chris R / Clear1nsight (#4546/#4543), Ryanson Jonathan (#4567), Cinna Davis (#4470), zxv, Jacob Alan Sussman and the retained scalar/Clifford/geometry/outer contributors, plus the concurrent investigations above. Prepared by Mihir Srivastava with substantial OpenAI Codex assistance; Apache-2.0. This is author-checked evidence, not independent human or formal-kernel verification.

[mihir-s-05]

Hosted verification status for exact head 23c515a3c071a6bfdd1fe80bc67cea55f6d43643: run 39745665141 stops in the shared onboarding tests; this package's hosted replay is skipped, not passed.

All four failures are in tests/test_research_current_status.py:

• CI navigation still names make frontier1094-verify, while the selected record requires make frontier4359-verify.
• The current-status test expects the exact bold token **0.0023**, but the existing page has **0.0023 = 23/10000**.
• The expected historical snapshot banner is absent.
• The inherited heading ## Historical selected status through #1094 violates the heading check.

I reproduced exactly four failures with python -B tests/test_research_current_status.py (six tests total). The test and all three inputs have identical Git blobs in this PR and base f626db5c3a3d781d8f7695038776c621651f75e1:

Path	Identical base/head blob
tests/test_research_current_status.py	0290040de1b2440649fda36b58e23f59b4e23d63
docs/ci-verification.md	582e416aaf1a8b5d40be80534bb069f4493a2433
docs/research/current-status.md	6e53bdc2654d6c3c798ade78de3240b9c56e02de
certificates/selected-result.json	ed33caef5c5827693e9572d252a66f99cbb73d40

These are existing base-documentation failures; this PR only adds its scoped package. The package's registered local replay, optimized semantic controls, metadata/source pins and exact committed-byte check pass as recorded above. No hosted package replay or maintainer scientific acceptance is claimed.