Make capability reindex deploy step best-effort

[cursor]

Summary

• Make the production deploy capability-vector reindex step best-effort after the healthcheck passes.
• Capture HTTP status, curl exit code, and response body when the maintenance POST fails.
• Emit a GitHub Actions warning instead of failing the deploy on reindex HTTP/network errors.

Root cause

The production deploy failed in 🧭 Reindex capability vectors (Vectorize) because curl --fail received HTTP 500 from /__maintenance/reindex-capabilities and exited 22. This post-deploy refresh is non-critical; the strict deployment gate remains the /health check immediately before it.

Validation

• Targeted harness against the exact workflow run block:
  • HTTP 500 returns exit 0 with ::warning title=Capability reindex failed and response body.
  • curl exit 22 returns exit 0 with warning and response body.
  • HTTP 200 reports Capability reindex completed (HTTP 200).
• npm run validate passed: format, lint, typecheck, unit tests, Playwright E2E, and MCP E2E all exited 0.

  

[kody-bot]

Closing as superseded by #625 (merged) — same rationale as #620: reindex is fixed at the root rather than made non-blocking.