It’s suspected to be related to interrupted backups, including interruption caused by intermittent networks. There are recently test cases, and bugs filed, so this is probably waiting for a developer to look at what’s going on. Reproducible problems are a fine first step though. There’s much less hope of fix without them.
“Unexpected difference in fileset” errors [2.0.3.9]
Failed put operations do not result in error #3673
It’s not really a missing file, but an internal difference between ways to count files that should be the same. With enough SQL queries, one could probably derive better information, but I don’t have the query offhand. Knowing the file is just the first step. After that would come getting things in agreement (or try the deletion).
It’s unclear what all the causes might be, but avoiding interruptions by things like sleep or reboot may help, and network issues can sometimes be avoided with a larger –number-of-retries. You can see whether it’s been doing retries recently by checking Job → Show log → Remote to check for put → rename → put sequences which imply retries. Click on the lines to see Size and Hash info to be more sure it was a retry. For longer-term unattended logging, you can add –log-file and –log-file-log-level=retry to your job or server.