Hi Railway team,

My service api.predictiqpro.com (project: natural-mercy) is returning 502 "Application failed to respond" on every request, with a suspiciously exact and unmoving pattern I can't explain from my own code.

Symptom:

Every HTTP request — across multiple unrelated routes (/fixtures/, /predictions/live) — fails at exactly 15 seconds, every single time, no variance.

The deployment shows status Active, not Crashed.

Build Logs complete successfully.

What I've ruled out:

I added explicit socket timeouts (3s) to my Redis client (redis-py, connecting via REDIS_URL) — no change to the 15s failure time.

I added explicit client timeouts (10s) to my Supabase client — no change to the 15s failure time.

I swapped REDIS_URL from the private redis.railway.internal hostname to the public proxy URL — no change to the 15s failure time.

Since none of these internal timeout changes affected the number at all, it doesn't appear the delay is coming from inside my application code.

The part I can't diagnose further myself:

Deploy Logs for the Active deployment show "No logs in this time range," even though PYTHONUNBUFFERED=1 is set and the app should be printing startup output (it did on a previous deployment, and does on a separate service in the same project running an MCP server).

I can't tell whether my app is even receiving these requests, or whether something is timing out before reaching it.

What I'm asking:

Can you check whether there's a platform-level (gateway/proxy) timeout of ~15s being applied to this service, separate from anything in my app?

Why are Deploy Logs empty for this deployment despite it showing Active?

Reference:

Service: api.predictiqpro.com (service name IQ-pro- in project natural-mercy)

Recent deployment IDs showing this behavior: 5b7bfff5, 3bc8ab64, 27e92861

Example failing requests: 2026-08-02 18:20:40 GET /predictions/live → 502, 15s and 2026-08-02 18:21:06 GET /predictions/live → 502, 15s

Thanks for any help digging into this from the platform side.