dump with -L flag - machine truely unresponsive

[Freminlins]

I have a FreeBSD 5.4 machine with a SCSI hard drive. I dump all our
machines onto a nfs-mounted filer. There are some files which it's
important not to clobber during the dump, so I've been experimenting
with dump using the -L flag like so: dump auLf /var /mnt/dump/var

What I have noticed is that during the time that the machine is taking
its snapshot, it becomes completely unreachable on the net. It is
impossible to ssh or even ping it. Our monitoring reports that the
machine has dropped off the net, but once the snapshot is finished it
is available again. The machine itself does not drop, it just appears
to drop.

Has anyone seen this before? I certainly haven't. And if so, does
anyone have a fix (apart from going back to dump without -L)?

For info the partitions are suspended during the snapshot for these
amounts of time:
Jun 21 03:00:08 gia kernel: /: suspended 7.003 sec, redo 0 of 4
Jun 21 03:08:41 gia kernel: /var: suspended 57.036 sec, redo 25 of 2775
Jun 21 03:22:19 gia kernel: /home: suspended 1.037 sec, redo 0 of 4
Jun 21 03:23:56 gia kernel: /usr: suspended 40.246 sec, redo 0 of 14


Thanks,
Frem.