The clusterpurge script and cronjob can delete various data stored on a cluster. What it can't do is to clear up broken cluster data:
- ezdfsfile rows for which the file has been deleted
- files for which the ezdfsfile row has been deleted
Both are worthless to keep around, and should be deleted.
How they appear is unknown and may have various causes, one possibility is PHP crashes during cache building, or timeouts on file operations. However, they seem to accumulate on some sites, and should be dealt with.
The existing "cluster_maintenance" cronjob part is a suitable place for this. EDIT: A bin/php CLI script would be better.