[vincentkoc]

What Problem This Solves

Review-sync preservation checks could pass failed SQL reads by reusing old values, comparing two default counts, or ignoring a snapshot iteration error.

User Impact

Test-only cleanup. Review syncing and archived data handling are unchanged.

Why This Change Was Made

Remove two redundant identity readbacks, JSON parsing and fallback handling; use the same PR8/R1 identities explicitly written by the fixture and always run wrong-node rejection. Check the seven remaining assertion queries and snapshot column/iteration errors. Preserve the six canonical snapshot tables, partial-response and cancellation cases, repository rejection and transactional rollback assertions. Record the cleanup in the test ledger.

Evidence

• Eighteen race-enabled before/after proof cases against temporary SQLite archives. Healthy cases pass in both versions; five failed-read scenarios pass before the repair and fail afterward.
• The snapshot probe reads a real row before SQLite raises an iteration overflow, exercising the missing error check.
• Positive-content, disabled node-validation and canonical-data-write controls all fail in both versions, confirming retained regression coverage.
• Module tidiness, formatting, vet, vulnerability scan and deadcode checks pass.
• Independent P0–P3 review: scoped-clean.
• Platform CI is required before landing. Overlays are disposable, and the PR changes no production code or live archives. No releases.

[clawsweeper]

🦞👀
ClawSweeper picked this up.

Pull request received. I will update this pull request when review starts.