fix(db): pin the primary-key plan in the search retirement read-bound test

[waleedlatif1]

Summary

• Fix the flaky bounds the IDs each page reads by the row limit once the limit shrinks test in 0027_retire_search_embeddings.integration.ts (failing staging CI with expected 199619 to be less than 90000)
• Root cause: the test disabled only sequential scans, so on its small fixture table the planner could still answer each page with a bitmap scan that reads every remaining row. Which plan it picked depended on whether autovacuum had analyzed the freshly inserted rows, so CI flipped between passing and a deterministic ~200k reads
• Fix: ANALYZE document and also SET enable_bitmapscan = off (reset afterwards), pinning the primary-key walk production uses. Test-only; the migration is unchanged

Type of Change

• Bug fix (test flake)

Testing

• Reproduced the mechanism locally: forcing the planner off the ordered index scan gives the same failure (213355 reads)
• Fixed test passes 5/5 repeated runs and the full packages/db integration suite (17 files) against a CI-equivalent pgvector pg17 database provisioned with db:push
• Mutation check: making the migration use a fixed 25,000-ID scan window still fails the test (213499 reads), so it keeps guarding the real regression

Checklist

• Code follows project style guidelines
• Self-reviewed my changes
• Tests added/updated and passing (new tests pass the test-audit authoring gate)
• No new warnings introduced
• I confirm that I have read and agree to the terms outlined in the Contributor License Agreement (CLA)

🤖 Generated with Claude Code

[vercel]

The latest updates on your projects. Learn more about Vercel for GitHub.

Project	Deployment	Actions	Updated
docs	Ready	Preview	Oct 2, 2026 3:01am UTC

[cubic-dev-ai]

No issues found across 1 file

Confidence score: 5/5

• Automated review surfaced no issues in the provided summaries.
• No files require special attention.

_{Re-trigger cubic}

[greptile-apps]

Confidence Score: 5/5

[Low risk] Test setup adjusts database query planner behavior.

The PR appears safe to merge.

Summary

The PR stabilizes the search-retirement read-bound integration test without changing the migration.

• Analyzes the fixture table and disables bitmap scans so the test measures the primary-key walk.
• Resets the added planner setting after the assertion.

_{Reviews (1) · Last reviewed commit: "fix(db): pin the primary-key plan in the..."}