Semantics for delete with a networked filesystem

Hello,
I’d like to get a little feedback on what the most commonly accepted way of
dealing with deletes in a network filesystem using the IFS kit.

Consider the case where I receive a setinfo that asks for a delete on close
for a directory.

1.) setinformation irp that asks for a delete on close
2.) I check across the network to see if directory is empty
3.) set fcb to delete on close
4.) some remote process puts file in the directory in question, I have no
way of being notified
5.) receive close irp, i ask the network to delete the dir, and it fails it,
but my user thinks it is deleted, since the setinfo returned a success

As far as I can tell, there are 3 potential ways of dealing with that

1.) Actually process the delete in the setinfo irp, and basically make it so
that the user cannot turn off the delete flag once they have set it
2.) If the delete on the directory fails when I do the close, try to rename
it, like NFS
3.) Silently fail

Thanks.
-Jeff