I want to investigate a planar pushing problem, but one issue I ran into is that frictional dynamics seems to be anisometric. My experiment is that I have a box with x,y,z slide joints and a hinge joint about the z axis on top of a plane. I control the torque on the hinge joint and I supply it with a constant torque. I then observe the change in yaw per frame, relative to the yaw of the block. In reality the change in yaw due to the constant control should not change with the yaw of the block, but it is strangely periodic in simulation. I suspect it has something to with alignment of gridded particles for calculating friction. I attached the simulation results and the minimum reproducible code (requires mujoco_py). Also this only happens to non-radially symmetric bodies; it doesn't happen with a cylinder (sanity check).