Railway breaking changes, updated weekly
Every verified Railway trap in Footgun: changes and behaviors that break code or give wrong results, newest first. Each one has the error you see, the fix, and the date it was last checked. The list updates when the weekly check adds or retires a trap. For the full cause and sources, open the trap, or see all Railway traps.
fg-0079· added 2026-09-30· last checked 2026-09-30
Railway log lines written in the same millisecond come back out of order
Symptom
Rows of output are shuffled
Fix
Number each line you plan to parse and reassemble by number.
fg-0073· added 2026-09-10· last checked 2026-09-10
New code is not live after 'railway up' because an older redeploy won
Symptom
Logs show the previous version's startup lines
Fix
Let variable and volume changes settle, then run one railway up on its own and confirm a log line that only the new code prints.
fg-0072· added 2026-08-15· last checked 2026-08-15
Deploying a Railway cron service does not run the job
Symptom
Deploy reports SUCCESS with no run in the logs
Fix
Test with railway run, which runs the command locally with the service's variables.
fg-0080· added 2026-08-01· last checked 2026-08-01· stale, may be out of date
Railway MCP lists variable names but redacts every value
Symptom
valuesRedacted in the list-variables response
Fix
Use the CLI from a linked directory when a value must be read or compared.
fg-0081· added 2026-07-31· last checked 2026-07-31· stale, may be out of date
Railway service loses its state file on every redeploy despite STATE_DIR=/data
Symptom
History and state are empty after each deploy
Fix
Create a volume mounted at that path and confirm the mount in the deploy log.
fg-0077· added 2026-07-31· last checked 2026-07-31· stale, may be out of date
Railway custom domain never gets a certificate after the CNAME was set
Symptom
Certificate warning on the custom domain
Fix
Copy the current CNAME target and the TXT record from the service's domain settings into DNS, and remove unused custom domains if you hit the cap.
fg-0071· added 2026-07-05· last checked 2026-07-05· stale, may be out of date
A RUN_NOW style variable reruns the job on every Railway redeploy
Symptom
The job runs and emails again after every deploy
Fix
Delete the flag right after the manual run. Audit services for leftover flags.
fg-0074· added 2026-07-05· last checked 2026-07-05· stale, may be out of date
Node service on Railway crashes with 'Premature close' or undici errors at boot
Symptom
Premature close on token or API fetches
Fix
Pin the Node version in package.json.
fg-0075· added 2026-06-12· last checked 2026-06-12· stale, may be out of date
Railway agent tool says the variable was updated, but the deployed value never changes
Symptom
Agent reports applied or staged for deployment
Fix
Set variables with the CLI and verify in the runtime log, not in the agent's reply.
Have your agent check Railway changes before it codes