I would discount opt3 straight away really - Mach has been going too long for there to be a bug which can cause this loss on such a simple procedure - its a facing cut, the simplest of code routines, probably only a few lines long.

My money would be to concentrate on 1 & 2

The big question is - is it a one-off or does it do it every run?

If its a one-off occurrence then it may never be found.

Craig, Davek0974, thank you both for helping.

I'm also skeptical about Mach 3 having such a bug, its been around for long, and why would it show-up after a year of regular operation on the same rig, a dedicated PC that receives MINIMAL maintenance and changes.

Vdc and I at PSU seem OK.  Motor couplers OK. Temp of one axis a bit high, to my taste.

Monitoring pulse trains on a 4 channel scope, pretty boring pattern, using its memory to keep as much data as possible. Figuring out a way to have the scope halt acquisition after a failure. 

Davek0974, two such occurrences in one year, on the same 2H - long job. 

short up-date: event did not reproduce thus far, pulse train and signals to the coils of the stepper OK.