fix(ci): make production smoke resilient to Render cold starts

[Ayush-Kumar0207]

Problem

The scheduled production smoke has been failing even though the Vercel frontend is healthy. Logs show the hosted Render backend returning transient 503 responses during wake-up, followed by 429 responses because curl --retry-all-errors retries too aggressively. Since the health step uses curl --fail, the job exits before Socket.IO, OAuth, or deployment-auth checks run.

Fix

• replace aggressive automatic retries on /api/health with an explicit cold-start allowance;
• poll at 15-second intervals for up to ~3 minutes;
• only accept the backend once /api/health returns HTTP 200 and { status: "ok" };
• preserve the existing structured /api/status validation and all later hosted-contract checks;
• still fail if the backend never becomes healthy, so this does not mask a real outage.

This change is limited to .github/workflows/production-smoke.yml.

[vercel]

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

Project	Deployment	Actions	Updated (UTC)
codeverse	Ready	Preview	Aug 21, 2026 2:42am