Well, a fudge/ workaround...
Slowing the datastream (Serial write) from the AVR to, say, 200mS, reduces the incidence of corrupt display and reduces the cpu load.
To reduce the cpu load further, if the terminal screen is is reduced in height to its minimum (80x2), then the cpu load can be reduced to circa 10% (still quite a bit?!)
Test case was Example - AnalogInOutSerial with a 200mS delay in the loop
It's a bit of a Kludge, and is treating the symptom, not the cause, but gets around the the problem for the most part for my needs at the moment
Perhaps someone with the skills could do a proper job