Node.js backend has ~12s TTFB on every request — proxy delay before Express receives the request

Question

Hi, my Express backend on Railway (Singapore, Hobby plan) shows \~12s TTFB on every single request including a /health endpoint with zero DB queries.

I added a timing middleware as the very first thing in Express and confirmed Express handles requests in 0ms once it receives them — both log lines appear at the same

Railway timestamp:

01:46:27.159807211Z $$2026-04-25T01:46:24.003Z$$ RECEIVED GET /health

01:46:27.159815646Z health reached

The 12 seconds is entirely in Railway's proxy/ingress layer before the request reaches Node.js.

\- Another Railway service I run (fuelscope, Singapore, same plan) responds normally

\- Issue persists after deleting and recreating the service

\- Same on Railway-generated domain and custom domain - Serverless is disabled

\- The only visible difference between my slow service and the fast one: Metal Build Environment is enabled (and cannot be turned off — the toggle reverts immediately)

Environment - TTFB  
Local - \~117ms  
Render (same code, same DB) - fast   
Railway - \~12,000ms

Is Metal Build Environment known to cause networking issues? Could this be a degraded host issue for this specific service?