When performing erase commands, I have noticed that the command exits about
two minutes into the erasing, no console logging is done, and then when the
drive actually completes, the NCR controller gives a SCSI phase error fixup
(CCB 0xdeadbeef already dequeued).  Then the bus locks up.
I assume that this is because of a hard-coded timeout.
Is this a correct analysis?
If so, where is the code located or how do I adjust it?

At first I thought this was an electrical problem.
It has caused me headaches for weeks.

Similarly, if an offline command causes a rewind and then an eject
if activity occurs on the bus between the rewind and eject, it
locks up.