FWIW you can reproduce the problem with this simplified code:
G64
F500
G1 X0.1
G1 A0.1 F500
G0 X0.2
M30
When this runs, the pulse stream for the rapid (G0 X0.2) is missing its accel ramp (which is why you get no movement and just a buzz). As an added bonus, CV fails to blend the two feed moves (G1) correctly. (The in-line feed causes a delay between the two moves).
However if you take out the F500 off the A move - it all works as you'd expect. (No delay AND you get a properly formed accel ramp on the rapid).
Alternatively you can run it in exact stop (G61) WITH the inline F500 and then the G0 DOES have an accel ramp. (but of course you loose blending).
The problem is a combination of CV, A moves (both rotary and linear) and in-line feeds.
Incidentally - replacing A with Y in the code above, in-line feeds still kill CV even between X and Y. I've never spotted this before because I've never had the need to change XY feed in-line.
It's not productive to fix Mach3 since it has limited lifetime anyway now that Mach4 is here.
It's actually the other way round.
Mach4 "is here" because it wasn't deemed productive to fix Mach3.