Ron Everett of Silicart visited on November 7th and 8th, 1988 to consult
with me on the prospects for curing the K processor boards' tendency to
die mysteriously at random places in the system code.  For the most part,
he spent his time examining clock signals for "noise" and signs of
ringing, and found several signals with undershoots of in excess of 2 V.

If I may attempt to paraphrase his recommendation, it was that the
undershoot observed was excessive, and that such undershoot might be a
cause of erratic behavior in the K processor board sets, and that we
should make an attempt to reduce the ringing on critical signals through
the installation of terminating resistors on one or two dozen of the
worst-ringing clocks (or other critical signals), with the expectation that
this will produce an improvement in the behavior of the boards.

I substantially agree with this recommendation.  I have spent several weeks
now in attempting to establish a pattern to the observed erroneous behavior
of several of these board sets which implies a common failure mechanism or
failing structure in the hardware.  I have been unable to discern any such
pattern, because: (a) the state of the system software and debugging tools
is such that a precise description of the circumstances of a particular
failure case is "impossible," or "requires too much work;" (b) given our
inability to precisely describe any particular failure case, it is impossible
to tell if a particular set of circumstances recurs in a subsequent failure,