This may or may not help however when we called in slow point in time
restores to IBM the workaround we got was to run the client with the
traceflag disablenqr.

This runs the restore using the old style restore protocol which is much
faster than the no-query restore protocol when restoring from filespaces
with large numbers of objects in. There's an IX-number APAR for this
somewhere.

The specifics in our case were AIX server, NT client, 100GB NT filesystem
with 'q occ' reporting millions of objects. Brought a PIT restore down from
4 hours to 1 hour.

Anythings worth a try when stuck..

Alan