[exim] Autoreply database on NFS

I'm looking at a rollout of Sieve vacation autoreply for a largish
organization, which has multiple Exim hosts in front of a Netapp.

The question is whether the reply-once database is NFS-safe. Tracing
through the code/docs/Google, it appears that the autoreply transport
will use tdb, and that tdb uses fcntl() locking, which should work (if
slowly) on NFS provided that locking support is available. Since
autoreply volume should be a drop in the bucket compared to normal mail
volume, performance implications shouldn't be an issue.

Has anyone been down this road ahead of me? Does it all work as
advertised? Any advice of which I should be aware? TIA.

--
Anthony de Boer