I turned on debugging in ser.c in effort to figure out why
serial isn't working anymore after the disk crash. All it revealed
that even after the rts/cts drops and nothing gets from/to the modem,
the kernel keeps happily doing serread() on it when running cu (which
seems to be busy-waiting on it). Doing a cat /dev/tty00 caused one
serread(), but the thing stayed up (I did this before cu). Generally 
it seems that something gets broked after a few calls to serread() and
the serial port happily closes and reopens when requested, but nothing
gets anywhere before the machine is rebooted. On an older aug 10 kernel
from puma it took a few minutes of cu and pppd before it took a walk
again.

-- 
'Good-bye and hello, as always'