using /dev/cua00 makes processes immortal?

[William Harold Newman <william_(_dot_)_newman_(_at_)_airmail_(_dot_)_net>]

I'm trying to set up and test serial communications over a null modem
cable in preparation for going to a computer Go tournament where that
kind of communication is expected.

I hung a null modem cable between my two old OpenBSD 2.7 boxes, a
Micron desktop and a Toshiba laptop. (The old boxes are the ones which
have serial ports, and I haven't bothered to upgrade them recently. My
new laptop runs 2.9 upgraded to the patch branch a few weeks ago, but
doesn't have a serial port.) When I ran
  # yes > /dev/cua00
one one machine and
  # cat < /dev/cua00
on the other machine as a first sanity test, everything went
swimmingly, with a stream of y's emerging from cat. However, now that
that sanity test is over, I'd like to kill the processes and do
something else. Unfortunately nothing works, not ^C, not "# kill", not
"# kill -9", not killing the shell that started them: both processes
("yes" and "cat") refuse to die. Rebooting both machines for every
test iteration doesn't seem like an ideal solution.

I did some Google'ing for "openbsd kill serial" and found a message
  <http://archives.neohapsis.com/archives/openbsd/2000-12/1623.html>
"OpenBSD - can't kill kermit process", which sounds like the same
problem. Unfortunately, I couldn't find any reply to that message.

Any suggestions?

(I'd be happy to supply more information if I knew what was relevant,
or willing to upgrade the boxes if I knew that it would solve the
problem.)

-- 
William Harold Newman <william_(_dot_)_newman_(_at_)_airmail_(_dot_)_net>
The King with half the East at heel is marched from lands of morning;
His fighters drink the rivers up, their shafts benight the air,
And he that stays will die for naught, and home there's no returning.
The Spartans on the sea-wet rock sat down and combed their hair.
   -- A.E. Housman
PGP key fingerprint 85 CE 1C BA 79 8D 51 8C  B9 25 FB EE E0 C3 E5 7C