502 dial timeout from BOTH the custom domain AND the .up.railway.app URL = railway's edge can't reach the container at all, regardless of routing. since you've confirmed binding to 0.0.0.0:8080 and target port is 8080, the gap is between "log line printed" and "socket actually accepting connections".

quickest split — railway shell into the service, then:

ss -tlnp | grep 8080
curl -v http://localhost:8080/

three outcomes that narrow it cleanly:

• ss shows :8080 LISTEN + local curl returns 200 → railway-side routing issue, tag @brody with project + service ids and these outputs as evidence
• ss shows nothing on 8080 → the app crashed silently after the startup log; the listen call never actually completed (uncaught exception in a deferred init step, missing dep, panic in a goroutine that took the process down)
• ss shows the listener but local curl hangs → bound but not accepting; init thread is blocked (deadlock, sync DB call, healthcheck thread holding the event loop)

once you know which branch you're in the fix is clear. without that split it's guesswork.

Hi Rich- I can't actually shell into the service. I'm just on the Hobby plan and I'm new to all this. Everything looks clean as far as I can tell. Should I delete the service from my git repo and re-add it and my env vars? Start from scratch?

Okay I re-added the Git Repo, re-set my env vars, a new domain with proper 8080 ports, and re-ran deployment. Network Flow Logs show only UDP and UNKNOWN protocol traffic — zero TCP connections are being routed to the container despite the app binding to 0.0.0.0:8080 and the target port being set to 8080. The proxy appears to not be routing TCP traffic to this deployment instance at all.

Forgot to hit reply, but I think I found something. Look at the previous post I just added here.

I have found the issue. It was deep in my code causing failures after build.