Render to Railway: what we learned migrating a production Rails API
https://xreplyai.com/blog/render-to-railway-migration-guide?utm_source=indiehackers&utm_medium=social&utm_campaign=blog-2026-08-05
Moved our whole backend to a new host with 1,558 scheduled posts sitting in the queue. The rule that made it survivable: suspend the old services, never delete them, because that is your entire rollback plan until the new database takes its first write.
Suspend, never delete, is a rule more migrations should follow, since the temptation to clean up early is strongest right when there is the least evidence the new environment actually works. With over a thousand scheduled posts in the queue, the trickier part is usually not the database write itself but making sure a job does not fire twice if both old and new workers are briefly alive during the cutover window. Did you pause the job processor entirely during the switch, or let it keep running and rely on idempotency checks to catch duplicates.