machine: 164lx with 2 21040 cards running 1.5.1  de driver

Both Ethernets were working okay until Friday, when one suddenly
stopped working.  Switched which network is connected to which
board, problem follows network, so board seems okay.  Swapped
out AUI cable, transceiver, measured coax (10base-2) with ohmmeter,
etc. etc. no joy.  Outgoing pings sometimes increase count as shown by
netstat -i, sometimes not?  In all cases Opkts = Oerrs.
The card *is* sending something out, and packets from remote nodes
*are* making it in, so the network hardware seems to all be working.

Sometimes get "de0: abnormal interrupt: receive process stopped"
is this anything to worry about?  This was only occuring on the
bad network until a few minutes ago when I also got it on the working
network, which continues to work fine despite the error message.

Looked through mailing list archives, found a message about de vs tlp.
Built a new kernel, but the tlp driver says:
"sopwith /netbsd: tlp1: sorry, unable to handle your board"
for the 2nd network, and tlp0 still doesn't work, although
netstat -i now reports a huge number of collisions, greater than the
number of packets (how can that be?), and no output errors,
so it is doing something different than the de driver.

Didn't see anything about improvements to the de or tlp drivers
in the 1.5.1 to 1.5.2 changes lists, so doesn't look like
updating to 1.5.2 would buy me anything.

I've added "debug" to the ifconfig argument lines in rc.conf, but
haven't seen any additional messages.  Do I need anything in the
kernel config file to enable this?

So... as far as I can tell no hardware broke, and other than the
rare y2k type problem software normally doesn't break just sitting
there, so I'm going nuts trying to debug this problem.  Any ideas?

-Snoopy