First of all, let me say that I didn't have time to thoroughly test this, and I might be wrong in my findings.
It seems that at least in the master branch when one deletes ALL bibdocs from a record (say record:97 in the Atlantis site) and runs a search with: filetype:a->z Results are returned even for that record, when they shouldn't. (Results are also returned if one searches the deleted filename in the filename index) It seems that my /normal/ bibindex task does not see that something has changed in the record. Even if one runs it manually (bibindex -a -i 97 -u admin), all indexes remain the same. If one forces reindexing of the specific record (bibindex -a -i 97 --force -u admin), ALL indexes are recomputed and everything works as expected, but this is not applicable to a production system. Should I use a different setting in my everyday bibindex tasklet, or is this a bug? Kind regards, Theodoros