Actually, I've seen something rather similar lately.  Again two
i386's, one a 486 running 1.3.2 as an NFS server (WD 8013WC, we0), the
other a PP40 running 1.3E (few days old) as the client (Tulip, de0). 
Occasionally the nfs connection would hang, and I'd be informed that the
host wasn't responding.  I'm on a hub with no other machines running at
the time (no uplink, either).
Tried pinging the 486 from the PPro, no response.  Tried pinging the
PPro from the 486, and the nfs connection instantly woke up.  When I
stopped the pinging, the nfs connection hung shortly afterwards.  I
didn't take much note of it (just kept the ping running to avoid further
problems).  I had never seen this problem before (and hadn't changed the
setup on these machines several days beforehand), and haven't seen it
since (though I installed -current on the 486 shortly afterwards too).

    Just a bad day for networking and NetBSD I guess...  =)

                                                          Will Ferry

------------------------------------------------------------------------
 William O Ferry  <woferry@CMU.EDU> | finger: woferry@Light.RES.CMU.EDU
 http://light.res.cmu.edu/~woferry/ | talk:   finger for online status
------------------------------------------------------------------------