Thanks.
Just today, all Machine VMs (of app, udns) went down at the same time. I want to move production traffic to Fly over from Cloudflare by Dec (because Machines might be 2x cheaper for our workloads despite the egress fees levied by Fly)… But: Machines going down all at once like a pack of cards is scary for a DNS service that pretty much needs to be available 24x7. I’ve left those unresponsive Machines intact for now (all stuck in the started state since forever) in case Fly needs them around to debug.