Long thin spinning objects take too long to stop spinning

See the SpinningObjectStability test map in the Physics Sandbox. [Link Removed]

It's not clear if this is a bug or a side effect of some of our solver shortcuts (e.g., omission of the precession term in integration, or PBD-style friction).

The problem can be avoided by adding some angular damping to the object, or increasing the inertia to make it less asymmetric, although both of these options affect other aspects of the object behaviour.

Create a capsule in a scene scaled by (0.2, 0.2, 1), and rotate it by 50degs about X. Then set the angular velocity to (0,0,1000) deg/s. The capsule will behave like a spinning top, but stays spinning for much longer than it should. It seems to be able to maintain a much lower angle with the ground than should be physically possible, as if the inertia is much higher than it should be, or the center of mass much lower than it actually is.