Don - did you ever resolve this issue?
I have been struggling with almost the exact same issue in the same application - a multi-purpose bridge crane used for a carbon anode baking furnace.
PowerFlex700VC 125A drive drawing from a common DC bus supplying the bridge, trolley, and anode hoist drives. The bridge drive runs two 20hp Reliance RL-model bridge motors controlled by a ControlLogix PLC via ControlNet, all engineered by Rockwell and installed new. No encoder on either motor. The drive nor the individual motor overloads never trip. Been a problem every since the installation in 2007. Rockwell field engineers have been in four times to no avail.
The local Rockwell distributor has said he'd like us to go ahead and try another manufacturer's drive to see if that solves the problem. Say what???
Currently running in V/Hz mode, with the speed regulator just on the verge of being turned off - seems to be the best solution, but nowhere near acceptable. If I'm not ready when the operator takes off, the repetitive jerking occuring on about a 1/2-second frequency interval just about knocks me off my feet. The operators refer to this crane as the bucking bronco.
An identical sister crane has the A-B 1395 DC drives and another larger crane of similar function has the Reliance GV3000 AC drives - both ride like Cadillacs. The problem-child crane used to run well with its 1395 drives, until the conversion. Of course, the drives that have been proven to work well are now obsoleted.
The jerking seems to be the worst when taking off from a stop, trying to position at slow speeds. Wide-open throttle take-offs aren't too bad but not great either. Joystick verified for clean signal. Motor brakes verified for full release. Several parameters such as jerk rate, accel/decel rates, current limits, speed regulator gains and the like have been run from one extreme to the other to find a happy/happier medium. Once things are moving at, say, greater than 10-15Hz, it runs smooth. Decel is pretty good until just before zero speed.
The amperage trace on accel shows the fluxing current spike on start-up/energization, then what appears to be a upright half-wave rectified pattern starting out ranging from 0.5 minimum to 1.0 maximum of the drive FLA limit, then tapering off to a relatively steady current draw in 3-5 seconds.
Would appreciate any suggestions, as my magic hat is out of rabbits.