I did some more troubleshooting last night, but I think I am even more confused now than before.
Tried the OF command. It did not seem to have any effect. I moved the cable to the E-H connector, so the X axis is wired to E and Z is wired to G. Same results as before.
If I disable the servo drives and turn the motor by hand, I get +2000 counts/rev in the CCW direction and -2000 in the CW direction, which is exactly what it should be. I jumpered the encoder to channel H to see what it would read. It is exactly the same when turning by hand, but is running about -250 counts/rev when the motor is drifting. This is odd, since the motor is drifting CCW, the + direction.
If I disable the servos and turn the motor several turns by hand, the Galil position error will accumulate as it should. When I reenable the servos, the motor will turn the proper direction to return to position and will then resume the drift.
The encoder shaft is not slipping on the motor shaft, but I didn't check the encoder disk yet. The encoder pulses are a square wave on the scope. The pulse length does vary, but the drift is jerky so this is the way it should look. It is difficult to see the A/B phasing with the uneven pulses, but it looks to be correct.
When the motor is drifting, if I remove one of the encoder leads it has no effect on the drift. The counting on channel H will stop, but the motor continues in the same speed and direction. If I reverse the encoder leads (A&B) the motor runs away in the same direction as the drift.