(NOTE: Marked Critical Priority because this involves data loss.)
A few weeks ago our Artifactory server (version 4.7.7) ran out of disk space during a nightly incremental backup. As expected, this triggered emails to admins, who increased the disk space for the VM. Subsequent backups succeeded and backed up new files deployed since the previous backup.
However, this week we attempted to restore from that backup onto another VM (pre-upgrade test install of 5.4.6) and around 20 files failed to restore due to mismatched checksums. Further investigation revealed that the failed backup on the original VM had written some zero-byte files instead of the correct data. On subsequent nightly backups it had not corrected this error, I assume because the timestamp on the corrupt (zero-byte) files was still the same as the "live" files – they had not been modified since.
I was able to resolve the immediate problem by deleting the zero-byte files and re-running the backup, which re-created them with the correct size/content. However, as far as I know, the only way to detect this during normal use is to remember to run a manual scan for zero-byte files, after recovering from any case of "out of disk space".