>The key indicator is *always* the massive slowdown when the *second* user starts accessing any database files being used by the first user. The first user is seeing a "simulated" performance whereby all of the data is being cached locally. When the second user logs in, the opportunistic locking that is being used with the SMB network redirector is broken, and the performance reverts to the real performance, which is really bad to start with (~1 second for a single row deletion is horrible).
Tish and tush (this is a family newsgroup after all) I should have remembered that one