I know it can be a lot of trouble to do, but the way I try to isolate such a problem is to swap devices until I can narrow down which item the fault seems to "travel with".
Eg: swap motors with a different axis to see if the problem moves with the motor.
Swap drive amps.
The controller could also have a fault, but if none of the above swapping seems to indicate any device as the culprit, then suspect something in one of your cards. The factory should be able to advise you once you have completed all these tests.
Shot in the dark: possibly retune the offending axis. You might have to check the documentation for your drive amps to locate if there is a balance pot on each amp. This setting can be tweaked to find a spot where the motor does not drift either direction. If I recall correctly (IIRC), this setting is done with the amp gain initially turned way down, and the command signal (+-10V) temporarily disconnected. The motor should not rotate without the command signal, and the balance is what sets this. The gain can then be turned back up to see if the setting holds. This test might be most meaningful if the motor is not attached to the load, to help remove friction from the equation.
Then there is the retuning procedure, but AFAIK, none of the PID settings actually will cause a motor to drift.
After this, then you should carry out the PID (proportional, integral and derivative gains) retuning procedures. Compare the current parameters for this motor's PID with the settings being used for the other motors. Similar motors driving similar loads should have similar PID parameters.
Just going from memory here (I'm at home right now):
Integral gain is very touchy, and should have a conservative setting from 0 to 5.
Proportional gain is not so touchy, and could have a range somewhere from 25 to 100.
Differential gain is least touchy, and could have a range from 500 to 2000