We have a four node cluster and find that when manually migrating resources
we can see unload times of 45secs to 1 minute resulting in users losing
connections to documents on network drives and causing a lot of problems.
We think that also when a node fails the reload time might be better but
the problems of lost connections as we say below is still there.

We hear that an NSS should unload in under 5 seconds and then reload to
avoid the loss of connection?

What can we check to improve performance - are there settings at the
cluster and/or on the client. At the moment it's not really helping much as
when a user loses connection MS Word seems to be reluctant to allow the
user to save the work elsewhere and often it's a case of them rebooting
their PC and us having to try and find temp files on their PC or using
salvage on the server.