If you ever find yourself trying to construct a triangulated representation of a model with quads, at some point you'll run across the problem of which vertices to link together to form triangles.
In the diagram above, two such orders are shown:
- (1,2,3) + (2,3,4) on the left does not accurately represent the quad, resulting in a hole in the side + a nasty overlap
- (1,2,3) + (1,3,4) on the right is what you actually need
I mention this because Genscher checking out some vertex order bugs the other day with some physics sims (IIRC it was with fluid/cloth). Double-checking what order the rigidbody stuff I've been working on was using, we found that it had basically been doing it wrong! I've since rectified this issue in the branch, though this still doesn't seem to fix the box+plane case where both objects are meshes!