Weaker joint drives in PhysX 3.3.3 when compared to 3.2.0

I’m simulating a hydraulic excavator in PhysX 3.2.0. The simulation models hydraulic rams as 2 bodies prismatically driven using a D6 joint (it’s loaded from repx so everything is D6Joints) with a damper, target velocity and limited force. I’ve configured the part masses to attempt to maintain a no more than 10:1 ratio between connected bodies. Arguably I should have modeled my system as an articulation but there are a lot of ‘loops’ and I believe at the time PxArticulation was still experimental.

I’m now attempting to upgrade my simulation to 3.3.3. Having dealt with the various interface changes I’ve got my system running but I’m finding that a lot of my hydraulic rams are weaker than they previously were (by a factor of about 6). Why would this be happening?

I’ve experimented with increasing the force limit but this doesn’t seem to make them any stronger.
The target velocities are not particularly small.
I’ve experimented with eDEPRECATED_32_COMPATIBILITY (Which I gather only applies to hard limits) and eDRIVE_LIMITS_ARE_FORCES which should only change things by a factor of 120 for my 120Hz simulation.
I’ve tried upgrading to PhysX 3.2.5 which seems to work a lot more similarly to 3.2.0.

At the moment my conclusion is that 3.2.0 can solve the system I have set up reasonably accurately and 3.3.3 cannot. This seems like a pretty significant regression, however, so I’m wondering if I’ve missed something.