Project ID: c2472429-b3ca-40e7-81ef-59437f042a05

Service: focused-radiance

Region: us-west2 Started: 2026-04-27 15:21 UTC, still failing 2h+ in

Outbound TCP from my service to api.ebay.com on port 443 is timing out at the connect layer. Other outbound traffic from the same service is healthy (Discord webhooks, our own Railway Postgres in the same region, public web traffic on the custom domain).

Error from Node's undici fetch:

TypeError: fetch failed

cause: UND_ERR_CONNECT_TIMEOUT

Connect Timeout Error

(attempted addresses: 66.211.166.2:443, 209.140.129.1:443, 209.140.129.24:443, timeout: 10000ms)

Independent reachability check (UK home connection):

-TCP connect to all three IPs: succeeds in ~150ms

Full HTTPS handshake to api.ebay.com via either IP: ~500ms, returns 401 (expected for invalid creds)

So eBay's API is healthy globally — only Railway egress from us-west2 to those specific IPs can't establish a connection.

Things I've tried (none helped):

• Service restart from dashboard
• Forcing Cloudflare DNS via NODE_OPTIONS=--dns-result-order=ipv4first (Node still resolves to the same IPs, so the issue is below DNS)
• Bumped the OAuth fetch timeout from 5s to 15s (undici's own 10s connect timeout fires first regardless)

Same issue from us-east4 ~14h ago: another user's thread - "Had this issue with ebay connection" by stevgarz — identical symptom, same eBay IP ranges (66.211.x, 209.140.x), cleared after ~1 hour for them. Mine is now 2h+.

This is reproducible across two Railway regions in <24h, looks like an egress-routing pattern rather than a one-off bad node.

Could someone check whether outbound from us-west2 to those eBay ranges is being dropped, or our egress IPs are filtered upstream?