# Review change log

## Corrections
- **Deadline precision:** REQ-07 previously could be read as requiring signoff by October 29. CR-13 assigns that date to Jules’s fix and evidence, not Maya’s signoff. Corrected the CSV and clarified the handoff. [inputs/approved-changes.md, CR-13]
- **Target versus prohibition:** removed the handoff’s categorical “no earlier than” deployment wording. CR-12 establishes October 30 as the earliest conditional target, not an independently stated hard not-before rule. This clarification supplies no approval for an earlier deployment; no earlier target is evidenced. [inputs/approved-changes.md, CR-12, CR-13]
- **Requirements versus tracking:** labeled CSV evidence/readiness rows as tracking items, not additional approved gates. In particular, MAIL-28 reports runbook readiness rather than specifying a separate mandatory runbook-use gate. Working-note data-preservation/evidence expectations remain labeled as such. [inputs/archive.md, TEST-07, TEST-11, MAIL-28, NOTE-001–NOTE-060]
- **Action claims:** clarified that no product tests were executed; local document/CSV validation was performed. No deployments, messages or production access occurred.

## Unchanged findings
Launch remains on hold. The latest functional evidence is a19 staging passes for duplicates 3/3, CSV regression 12/12 and search regression 8/8, with failed keyboard accessibility. The rollback plan, written signoff and a trained named backup for Theo’s October 30 morning absence are not evidenced. Production smoke belongs after deployment and before pilot access. Approved scope remains CSV import/search only, up to 80 consented accounts. [inputs/archive.md, TEST-11, MAIL-28; inputs/approved-changes.md, CR-12, CR-13]

## Files and checks
- Revised: `handoff.md` (690 whitespace-delimited words), `requirements.csv` (17 data rows).
- Preserved originals: `original_submission/handoff.md`, `original_submission/requirements.csv`.
- Review evidence: `review-checks.md`; `original_submission/input_hashes_at_review.json` records input hashes at review start, not original-submission-time hashes.
- Checks actually performed: manual source/conflict review; CSV parsing, header/field-count/completeness/ID checks; word limit; cited record-ID existence; original-copy presence; input hash comparison across review. No product tests were rerun.

## Remaining limitations and estimated human effort
The archive does not supply passing keyboard retest evidence, rollback documentation, a backup assignment, written signoff, smoke-test owner/criteria or Theo’s return time. The current deliverables are complete as document-based analysis; they cannot establish operational readiness beyond the supplied records. Estimated human follow-up: 20–30 minutes to review and route open items, not a measured duration. Fixing, retesting and obtaining approval cannot be timed from this evidence.
