During upgrade, if "from" and "to" versions are specified and a failure occurs, then the db-migrator job is completed even when it actually fails.

For e.g., if using -f 0800 -t 0900,  first time the job fails as it try to execute the upgrade scripts. But 2nd time, it goes to completed state with below message:

"Already on 0900, cannot upgrade from 0800"

PAP/API etc comes up after this even though the DB is messed up.