The routing layer being down while workers/data services stay up is such a specific failure mode. Usually means the load balancers or edge routing got corrupted somehow, not the actual compute infrastructure.

If you're serious about migrating off (and not just saying it in the heat of the moment), the main thing is having a plan for the database migration. That's always the painful part. Everything else is just Docker containers that run anywhere.