Larger moving parts have a much bigger physics impact compared to moving smaller parts, which is more evident with complex collisions. If you instead take the mesh or part, converts it to a special mesh and places it into a really small part (0.1, 0.1, 0.1 for example), you can have the same large mesh but without the physics impact. See the video bellow for a representation of what i mean.
See how as i duplicate both parts, the larger cube lags behind my mouse a decent bit, while the small part has nearly no difference.
I see this useful personally for planes or any moving vehicle, as this can be used to simplify the physics geometry for large parts (Box collisions still have a physics impact because the collision hitbox is so large).
I made a plugin that lets you convert big parts to small parts with scaled meshes in them here: