I saw this happen with passdb backend = smbpasswd: the listing of users via pdbedit -L stops immediately if there are users in smbpasswd that no longer exist as POSIX users in /etc/passwd. That is very confusing and misleading. I don't have such a setup here right now, therefore no logs so far.
It stops with a message that the smbpasswd database is corrupt, which is true as it is no longer in sync with the actual users on the system. What exact behaviour do you expect?
it should not print some accounts, then issue an error and exit. If you fix that single reported error, you get the next one and so on. That might be a fun game for pdbedit but not for the user if the number of mismatches is >> :-) better approach will be that pdbedit issues a warning on each mismatch and eventually finally returns with an error.
Björn, the attachment 1 [details] in bso#7543 is doing exactly this, however there we are talking about smbpasswd migration and I am currently unsure if those accounts should be migrated or not.