We don't see things drop out for no reason. Ours give a reason (search the
server activity log for the node name that has dropped) of "session
terminated". What we do not know is why things terminate!

We believe our errors indicate a network problem. We have reduced the number
of these drop outs dramatically by:
 - spreading the backups out over a longer period over night,
 - moving "big files and fast nodes" (12.5 GB backed up in 24 mins) to a
time that does not clash with "small files, slow servers"
 - Changing CommTimeOut and IdleTimeOut values.
 - Insisting all network ports involved are tied at 100 Full duplex.

Cheers, Chris