Hi Railway team,

I have three custom domains on my project that all return {"status":"error","code":404,"message":"Application not found"} from Railway's edge, despite the Railway dashboard/API showing certificateStatus: VALID and successful deployments.

Project details

Project ID: 521bcef6-83f9-4085-8b98-86a7969e51ea

Environment ID: 256a9e5e-25fb-41b0-b7b0-1201cf33439f

Project name: Shlink Short Links

Affected custom domains

admin.goproactive.uk → Caddy service (ID 3d72d4b5-fe3a-44e3-96b7-bd6c87e18d31), targetPort 8080

links.goproactive.uk → Shlink service (ID 2c1bf774-b8c1-4e9a-a506-023bbeacbb3f), targetPort 8080

goproactive.uk (apex, via Cloudflare CNAME flattening) → same Shlink service, targetPort 8080

Recent request IDs (all returned Application not found from Railway edge)

XBpwv1MOT66vT5tSn6XIxQ (goproactive.uk)

-7IahYPRQmK0jufy-_9nXA (admin.goproactive.uk)

qzZAgJeLT-W6MKrc2prcFg (links.goproactive.uk)

Symptoms

TLS handshake succeeds; Railway serves the correct per-hostname cert (verified subject=CN=admin.goproactive.uk and subject=CN=links.goproactive.uk via openssl)

Cert status via GraphQL: CERTIFICATE_STATUS_TYPE_VALID

But every HTTP request to the hostname returns the "Application not found" JSON error

Requests to the underlying Railway-generated URLs (caddy-auth-production-3e09.up.railway.app, shlink-production-a692.up.railway.app) work perfectly — services are healthy

Behaviour indicates the edge has TLS termination for these hostnames but is missing the hostname→service routing entry

Suspicious signal

Deployment records show staticUrl: "admin.goproactive.uk" and staticUrl: "links.goproactive.uk" — I'd expect staticUrl to hold the Railway-generated *.up.railway.app domain, not the custom domain. This may indicate corrupted internal state.

What I've tried (unsuccessful)

customDomainIssueCertificate mutation on all three IDs (certs became VALID but edge still 404s)

Delete + re-create the custom domains (which rotated to new required CNAME targets; DNS at DNS provider updated accordingly)

Multiple serviceInstanceRedeploy cycles on both services

customDomainUpdate to explicitly set targetPort: 8080

Full DNS migration from Heart Internet to Cloudflare (nameservers now anastasia.ns.cloudflare.com / brian.ns.cloudflare.com; all Railway-required CNAME targets present, DNS-only/grey-cloud)

DNS verification

Cloudflare authoritative NS returns correct records:

admin.goproactive.uk CNAME → ac833s3s.up.railway.app

links.goproactive.uk CNAME → fymcktz6.up.railway.app

goproactive.uk (apex, CNAME-flattened) → pighjxu5.up.railway.app resolves to 69.46.46.71

Railway API shows dns_status: PROPAGATED for admin and links; apex shows REQUIRES_UPDATE (likely because CNAME-flattening returns an A record — a known limitation with apex-on-Cloudflare + Railway custom domains, but the underlying resolution is correct and the cert would be issuable via HTTP-01 challenge)

Request

Could you please clear whatever internal state is preventing edge routing for these three domains? If the fix involves rebuilding the domain entries entirely, please advise and I can delete/re-add — but I've already done that twice with no change, so I suspect there's project-scoped state that needs manual intervention.

Happy to provide any additional info. Thanks in advance.