Per my understanding, the error (Starting clean up) you’re seeing means either the main (entrypoint) app process couldn’t be started (that’d mean /cnb/process/web couldn’t be run as user heroku) or health-checks (as defined in your .toml) failed.
Other reasons I’ve seen of late:
-
Regions: Did you recently change regions? At least one user reported that changing regions midway a deployment procedure caused a failure similar to this: Fly clustering and scaling no longer working - #6 by mmark
-
Zombie: Someone had a zombie app instance block deploys: App instance stuck - #8 by miles What does
fly status --all -a <app-name>say, for your app?