My custom domain verifies fully (DNS + TXT + Let's Encrypt cert) but the edge has no route mapped to my deployment, so every request returns the fallback "Application not found". This

looks identical to a recent staff-resolved case (details at the bottom).

PROJECT / SERVICE

• Project: agentplatoon (ID 7a070d68-c79c-417d-aeb9-1193dd51050f)
• Environment: production
• Service: web (Next.js standalone, binds 0.0.0.0:3000, EXPOSE 3000)
• Custom domain: agentplatoon.com, target port 3000
• Railway-assigned target for the domain: 7zn2pw5s.up.railway.app

THE SERVICE IS HEALTHY — its auto-domain works:

• https://web-production-abf87.up.railway.app/ -> 200 (incl. /api/health -> 200)

THE CUSTOM DOMAIN + ITS ASSIGNED TARGET BOTH RETURN THE FALLBACK:

• https://agentplatoon.com/ -> 502, x-railway-fallback: true

• https://7zn2pw5s.up.railway.app/ -> 404, body:

  {"status":"error","code":404,"message":"Application not found"}

  on every path (/, /login, /api/health). A bound target never 404s.

TRACE PROVING IT'S A ROUTING-TABLE GAP, NOT NETWORK/TLS/EDGE HEALTH:

1. Network path is fine — timings to the failing target: DNS 21ms, connect 30ms, TLS 71ms, TTFB 97ms. It answers instantly with "Application not found" — nothing is timing out.
2. TLS/cert is correct — the failing domain serves its own valid cert (openssl -> CN=agentplatoon.com, Let's Encrypt). The cert subsystem provisioned the domain; only the route didn't.
3. The edge is healthy — agentplatoon.com / 7zn2pw5s resolve to edge IP 69.46.46.36. Forcing the KNOWN-GOOD host web-production-abf87 to that same .36 edge (curl --resolve) returns 200.

So .36 has working routes — it just has no entry for my custom domain.

1. Missing on every edge, not one stale node — forcing agentplatoon.com to the other edge 69.46.46.21 also returns x-railway-fallback 502. The route is absent cluster-wide

(control-plane), not stale on a single edge.

1. Not DNS/Cloudflare — Cloudflare CNAME -> 7zn2pw5s.up.railway.app is DNS-only (grey), no "proxy detected" warning in the Railway UI, and curl --resolve

agentplatoon.com:443: (bypassing Cloudflare) still returns the fallback.

DNS IS ALREADY EXACTLY WHAT RAILWAY ASKS FOR (pre-empting the usual first reply):

The domain's "DNS records" panel in the dashboard requests:

• CNAME agentplatoon.com -> 7zn2pw5s.up.railway.app — present in Cloudflare
• TXT _railway-verify -> railway-verify=cf252d3f0bac630308fd267f5788945b3f85d992e57afa8a4df38422c7389bdf — present in Cloudflare, byte-for-byte identical

So the _railway-verify TXT is correct and matches; the verification-token fix does not apply here.

ALREADY TRIED (no effect): reissue cert; delete + re-add the domain several times (each rotates to a new target hostname: dllg2nev -> llclf7ge -> scombc9j -> 7zn2pw5s, each 404s

standalone); multiple redeploys of the web service after the domain was verified + DNS-only, waiting 15+ min for edge-cache convergence; explicitly set target port 3000.

IDENTICAL RECENT CASE (staff-resolved): the Central Station thread "Custom domains return 'Application not found' from edge despite VALID cert status" (project "Shlink Short Links") —

same 404 body, cert VALID, generated *.up.railway.app URL works, delete/re-add rotates the CNAME target but never binds the route. Same fingerprint as mine, landing in the same

69.46.46.x edge range.

ASK: please MANUALLY CLEAR the corrupted internal routing state and re-insert the edge route mapping agentplatoon.com -> the active web deployment in production. The host->deployment

entry was never written to the routing table, and all self-serve steps above have failed. Suspected project-scoped corrupted state needing manual intervention.

Request IDs (all x-railway-fallback: true, edge lhr1):

e2bgPj6bSBy-4elDs_GTAg, b2IGXfWIQKO8sVVDw9P4nw, 9_2TyVlTRQmb_Yfl21mRUA (x-hikari-trace: lhr1.ppx8)