Repair historical CI inputs, generated hashes and regression fixtures

[antunishdPursuit]

Historical CI rejects an advanced main checkout when frozen packages still expect their original selected record, Makefile, or fused-block helper. Other regression checks need immutable Git objects absent from shallow checkout, later review batches, and certificates regenerated after performance integrations. This repairs those inputs and fixtures while retaining the original rejection conditions.

The final patch against current main c6765481:

• Authenticate three exact historical Git blobs against unchanged predecessor manifests and replay the unchanged R420, reconciliation92 and split-skip checkers in an isolated copy of tracked working bytes. The live Makefile must preserve its complete historical prefix. Only the declared historical input is replaced, and every tracked temporary file must retain its initial byte hash after replay.
• Keep the R420 native C++ checks, split-skip producer, independent arithmetic and mutation tests in CI. Frozen package files, source pins, upstream inputs and the selected frontier are unchanged.
• Regenerate stage-pair, rank-product-core and mixed-point certificates after the already-integrated performance implementations. Their only final differences are the three source-hash entries; mathematical fields remain identical.
• Fetch exact commits 0b1cccf86e2f5132ed286a26ef0a681a495740ea for performance regression tests and 407d52ff15b2217b6e22bce44728b4c3907b6515 for the Batch1047 historical inventory. No unpinned fallback or disabled check is added.
• Extend ledger fixtures through review batches 600–900 and emit POSIX historical references from the ledger generator on Windows. After syncing main and regenerating, the ledger artifact itself has no diff against current main; no review decisions, contributor identities or GitHub closure state are edited by this patch.

Validation:

• Hosted Linux at 8ee234cb: R420 on Python 3.11/3.13, including native C++ and reconciliation92 pass. All three clone jobs (Python 3.11/3.13/3.14) pass in the research-artifact run, including split-skip's original producer, witness, independent arithmetic and controls. That earlier run has 45 completed passing jobs; its six failed test/research jobs are the cases repaired by the follow-up commits. Final-head hosted CI remains pending.
• Windows CPython 3.11: all three audit generators pass; R420's replay and four tests pass; reconciliation92 authenticates 92 full trees and 4,564 distinct blobs; seven new isolation/pin/prefix/drift controls pass normally and with -O; six stage-pair performance, eleven rank-product, seven mixed-point, four inplace-side performance and fourteen reconciliation tests pass. The latter fourteen and all seven adapter tests were rerun after the main sync. The spectrum-audit failure was already fixed upstream (ten tests pass).
• A deliberately shallow disposable clone starts without both historical commits, then fetches them and verifies the exact performance certificate SHA and historical tree ID. All staged snapshots match their unchanged predecessor hashes; the final contribution diff passes whitespace checks; the checkout is clean.

Scope and limits:

This is CI maintenance, not a new multiplication result or a formal all-size theorem check. Final head 61532beb0701123a892bbef50179220ce9c6ed66 includes a main sync resolving the generated-ledger conflict without rewriting branch history. Current-main PR checks are pending; green status on earlier jobs is not final-head acceptance.

Older main failures were grouped across all 173 failed runs (11 workflow names). Missing PR721 logs are now present, and the Batch1047 missing-tree failure is addressed here. The historical #538: covered bytes changed: .gitattributes inventory gate remains outside this repair; five older workflow logs were unavailable. No claim that every historical main workflow is green is made. The unchanged worker suite needs make, and the frozen split-skip verifier uses POSIX path strings; their Windows runs are not counted as successful checks.

Coordination and handoff. Entry overlap pagination covered 367 open PRs and 651 Discussion records; matches were predecessor packages and credited stage-pair-performance work, with no active matching repair identified. Submission-time PR/board refresh found no new matching claim; metadata search is not proof of novelty.

Prepared by Dennis / antunishdPursuit with substantial OpenAI Codex assistance; Apache-2.0. Credit Scott Woolsey Biggart, Greer Guthrie, the reconciliation contributors, Chafik Boukhalfa, RaD/hipotures, Rohan Garg, and the original performance implementations. Exact snapshot origins and reproduction commands are retained in scripts/historical-ci-inputs/README.md.

Local commits: c530cc8a — fix: regenerate audits after stage-pair performance integration; 8ee234cb — fix: replay historical checks with authenticated isolated inputs; 7c6a72e6 — fix: refresh ledger projections through the latest review batches; a01d46be — fix: preserve portable paths in regenerated ledger references; 5fd04ce0 — fix: regenerate mixed-point audit after span-frame optimization; 27dda15c — fix: fetch exact historical Git inputs in shallow CI jobs; 61532beb — Merge current main and reconcile the generated review ledger.

[Mister-Meeseeks]

Batch review at 61532beb0701123a892bbef50179220ce9c6ed66.

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

Reviewed scope: Historical CI adapter authenticates unchanged predecessor pins, replays in isolated tracked-file copies and rejects changed tracked outputs. All three historical jobs, including R420 native checks, and isolation/corruption tests replay on the revised head using a dependency-complete bounded export. Additional changes preserve POSIX ledger paths, fetch immutable shallow-CI inputs and refresh three certificate provenance hashes only. Combined ledger tests are required before publication; no selected supplier change.

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.