Hi Railway,

platform_api in our pre-staging env has been unreachable via the public edge for 1+ hour. Every request returns 502 x-railway-fallback: true, but

the container is healthy — uvicorn bound to [::]:8000, deploy SUCCESS, outbound HTTP works. Zero ingress ever reaches it.

IDs

- Project: 2cb0b7c3-f5dc-4594-a2b0-05a32fe5ba65 (courageous-creativity)

- Env (pre-staging): 81c47819-783f-44fb-8a9a-fdf0400b236f

- Current service: cae4d821-c9b1-401b-87c8-2a86e2770f96, us-west1

- Public domain: platformapi-pre-staging-88dd.up.railway.app (target port 8000)

- Latest SUCCESS deploy: 7f74d1ce-f318-43f1-a902-4573ccfdb485 @ 2026-04-20T18:12:40Z

- Original service (deleted during recovery): a8f0f7af-df00-4b79-a1cf-02c7c4e9b153, was us-west2

Request IDs for tracing: f9ylfoIDS6aLJJ0umrpb1w, 0yqBo2kPSp2_iW4cipRofQ, hYoxz0l-TeKJZDhIV7rehQ, mv_f-S0lSWyQUvzV-8Y8hA.

App logs (healthy): Started server process [1] → Application startup complete. → Uvicorn running on http://[::]:8000, then successful outbound calls

to Twilio, ElevenLabs, Cartesia, OpenRouter. Egress works, ingress never arrives.

Recovery attempts (all failed, 502 persists after each):

1. deploymentRestart

2. deploymentRedeploy ×3

3. serviceDomainDelete + serviceDomainCreate (new URL, still 502)

4. Full serviceDelete + serviceCreate with 67 env vars preserved (new service ID, new URL, still 502)

5. Dockerfile BuildKit cache-mount IDs updated to new service ID, pushed as 160bc3d3 — cold build succeeded

6. Region us-west2 → us-west1 (Railway banner: "The deployment configuration was automatically modified to ignore deprecated regions")

7. serviceInstanceRedeploy — SUCCESS, still 502

Observations

- The us-west2 deprecation banner suggests that rollover likely initiated the wedge on the original service.

- Other pre-staging services in this project (FlowForge-Frontend, Tool_Hub, voice_gateway) route fine — no x-railway-fallback. Scoped to

platform_api's edge entry.

- Upstream routing appears stuck: every request short-circuits to the fallback before reaching the container. Held across full service recreation

with new ID, new domain, new region — rules out local config.

Ask: please inspect edge/upstream state for cae4d821-c9b1-401b-87c8-2a86e2770f96 in env 81c47819-783f-44fb-8a9a-fdf0400b236f and force

re-registration, or tell us the config step we're missing. Blocking login on pre-staging. Happy to keep the current deploy untouched while you

investigate.

Thanks,

Abdul Rasheed Esa