I've seen it too after trying it, but just once, it didn't reproduce well.
   Maybe you could do some packet sniffing and send me the output, I'll
   try to fix the bug RSN.

I've always thought that executing nfs rpcs during mount was a bad idea.
(I always had my code deferring things like nfsstat until the first I/O
 system call on the mount point instead of the mount system call, but others
 keep putting them back in:-) The problem is that it is difficult to avoid
race conditions (although I can't remember the scenario, its been a few
years and I'm gettin old:-) and is a hassle when the server is dead at
mount time. For a root mount, it just means "no boot", but for other mounts
it can result in cascading directory locks, since the ffs mount point is
locked during the nfs_mount() call during the mount syscall.

Just my $0.01 worth, rick