Ah, you know what? I tinkered again and found that with the card set as LPT1 and the 0xDC00 address in Mach3 it WILL move. Didn't seem to work with the card and Mach3 set to LPT2 even though Mach3 was given the 0xDC00 addr for LPT2.
Anyhow, guess what.. SAME PROBLEM.
I set the z-rate to 25 ipm. I can jog up and down pretty regularly at this speed. Set the zero up high, bring it low, tell it to "Goto Z" and in 5 sec like clockwork it stalls. Same thing for moving the X way out, speed =30ipm, one... two.. three... four... BUZZZ. Acceleration was like 4 or something, it wasn't so slow that it took 5 sec to get up to speed. It's up to speed in way under a second.
5 sec to stall from any position, exactly what I got with the onboard parallel. The consistency of its time-to-stall is pretty suspicious, as is the fact that it keyboard-jogs just fine. I did measure the keyboard jog speed and it seems consistent with what was programmed, as well as consistent with the speed demonstrated during the failed Goto Z or G-code executions.
I... really don't get this. Like I say I know there's resonance issues and that's an easy answer but it can't be resonance. It will stall at even lower speeds than this, and resonance would not take 5 sec to build up and do so with great consistency at 5 sec. Cable noise (~2 ft cable, and I've tried 2 different cables) wouldn't do this either, not with this kind of consistency. If I randomly lost a step or two and periodically stalled at random times I'd expect cable noise. I did build some dampeners earlier anyways, one a very interesting harmonic balancer design like a car, one by filling the knob with birdshot (pretty cool), another oak with steel rollers in slightly slack holes as often described. Makes no difference at all.