Hi! I was trying to weld some bodies together using a weld equality constraint. The weld works correctly in that it keeps the bodies together, however, it does not seem to simulate physics in the way that I expected. The specific example is with two bodies, each of them consisting of an elongated box geom. Each of the bodies only move within a 2-D plane with two slide joints, and a perpendicular rotate joint. Floor: ------- Body 1: XXXXXXX Body 2: OOOOOOO If I arrange the objects like in an inverted L shape with a weld between the two bodies: OXXXXXXX O O O O -------------------- The subsequent dynamics are just: OXXXXXXX O O O O -------------------- Essentially, the L does not tip over, towards the right, as it should. Is this expected? I have checked multiple things: - Masses, center of masses, and inertia tensors seem fine. - If I remove the weld, both objects fall all the way to the floor as expected. - If I add a point obstacle that hits the X body as it is falling, the entire body rotates/tips (so it is not a problem with a missing degree of freedom). - If the point obstacle hits the X body exactly in the middle as it falls, it also stays stable on the obstacle (Also unintuitive). - If I provide some initial rotation to the L towards the right side, it will still not tip over until the center of mass of the O body stops being supported (at which point it does indeed tip over). Any idea what may be happening here? Any suggestions for how to solve it? I know I could make just both geoms to be in the same body, or the second body as a child of the first without joints. However, for my use case, I want to be able to enable/disable mid- simulation. Thanks