Just for your amusement:

Our network at work has a large amount of traffic, from half a dozen busy VAX
clusters, to 20-30 unix boxes (no NetBSD unfortunately), a handful of PDPs,
a IBM mainframe and probably over 100 PCs, all on the same subnet.
Our newest PCs started having the exact problems descibed by others here.
Some just lost TCP, others hung. What's more, often 6 PCs would hang at the
same moment. BTW, these were running MS WorkGroups, not a real OS like
NetBSD ;) Our network group attached a monitoring tool to the net and
couldn't find anything. The solution? the Pentium 90s were upgraded
(downgraded??) to Win95, and the newest Pentium 100 PCs(already running 95)
were reclocked down to 75MHz. The problems didn't go away, but they were far
less frequent.

The final solution, yet to take place, is to replace all the ethernet cards,
and hope the problems disappear completely.

This won't help our NetBSD problems, but YOU ARE NOT ALONE!

--
   /\/\     Paul Ripke
  / / /\    BHP Information Technology
 / / /  \   VMS & UNIX (AIX, HP/UX, DG/UX, SCO, SGI, Digital...) Sysadmin
/ / / /\ \  Computer Centre, Five Islands Rd, Port Kembla, NSW 2505, AUSTRALIA
\ \/ / / /   ripke.paul.pr@bhp.com.au stix@itwol.bhp.com.au jpr36@uow.edu.au
 \  / / /          Anyone wishing to lay claim to the opinions expressed
  \/\/\/                      herein, do so at their own risk.