The tuning question came up before, but it's doubtful it's just that. Slight separation of the lines may happen, according to Zoof, the one person who ever tested this, but not that magnetic-like hysteresis that happens close to the zero-motion moment in either axis. It's a real killer, because obviously fine detail like the text in an ILDA pattern cannot be legible with a problem as severe as this, because motion must frequently stop in either axis when writing text.
About grounding, no problem there. I'm familiar with audio wiring, and was using a DC coupled balanced output. It was even tested for jitter on the ILDA pattern where the draw exists the circle and starts on the detail above it. If that bight, unblanked, does not jump about, then jitter is not a problem, nor is noise. (Pointing accuracy may be, but in smooth continuous motion, even WideMoves are ok with that.)
Edit:
I looked at the WideMove driver board, it's grounded to the baseplate via its mountings, as intended. I don't think isolation will help. (It's usually done to eliminate ground loops, whose result is slight sine-wave modulations of intended scan position, but that's a problem I've never seen with these). I'm certain that the warped-line problem is mechanical, because it dominates any motion that is either stopping, or trying to start, regardless of the signal strength. Another certainty is that if this were in the signals, then this system, usually used for audio, would act like a noise gate!It would mute all low level signals. That definitely never happens.