# Regression provenance

The supplied `followup-tests.log` reports 99/99 tests (two files), run at 09:44:49 UTC on 2026-10-21. The task brief, implement-followup report, and proof audit attribute this to exact production head [80d27e96db](https://git.cafe/pingdotgg/t3code/commit/80d27e96db888d3ab0dbf12314f136f4e8b3e9b4), full SHA `80d27e96db888d3ab0dbf12314f136f4e8b3e9b4`. This evidence step reused that supplied result as instructed. It did not rerun or relabel those tests as newly executed at the fixture commit.

R6–R10 reuse the committed recovery and client execution tests described in matrix.md: active supersession, root/first-input ordering, reset inclusivity and post-reset boundary, failure/root identity, and ordering/Stop/rearm/idempotence guards. The supplied 28-history SQL/shared parity diagnostic is supplemental, uncommitted diagnostic evidence, not a newly committed regression fixture.

The new fixture commit [9fd419a6e9](https://git.cafe/pingdotgg/t3code/commit/9fd419a6e9b455f17958325ad3ed696f08f00722), full SHA `9fd419a6e9b455f17958325ad3ed696f08f00722`, adds only three development script/fixture files. All six production/test files from the existing fix are byte-identical to 80d27e96db888d3ab0dbf12314f136f4e8b3e9b4. No test edits were made, so the 99-test suite was not rerun.

Fresh validation for the added scripts: targeted lint, formatting, server-scoped tsc-rs --noEmit (exit 0, existing suggestions), smoke captures, and paired actual-server captures. `capture-complexity.log` records a fresh benchmark at the full fixture commit, including nominal 3000-wake selector input, 3000-token malformed reset, and 3000-token nominal/malformed JSON parser input. Timing is diagnostic, not a timeout-based test.

No browser, UI, real provider quota, or Claude adapter exhaustion proof is claimed.