Now, let me start off by saying that I understand that this would require a major rework, and it goes against the fundamental theme of the game. But hear me out. BeamNG's soft-body physics system is really good for simulating anything that can bend and flex, such as metal, plastic, rubber, and so on. But what I've noticed is that the soft body physics system when trying to simulate something hard, like a rock or a non-deforming piece of wood (example: Cannon wheels), it is usually unrealistically bouncy no matter how the beam strength and damping is tuned, just because there's no way to make it truly solid. It'll always deform slightly and bounce(which is realistic, but at a significantly smaller scale). And, at least in the case of rocks or concrete barriers, they still deform like metal when enough force is applied. Granted, it takes a lot of force, but my point stands. And that point is that adding a system for proper hardbody physics would allow those non-deforming props to... Not deform. If a system like this was implemented, the props that use it would allow for direct tuning of mass, and bounciness, which are the two main things that the soft-body system currently makes deterministic based on node weights and beam properties. A better fracture system could also be implemented, even if the fractures are predetermined. So instead of a concrete barrier bouncing when dropped and possibly deforming unrealistically, it could actually break into multiple pieces. Same if a car runs into it at a high enough speed. If the two physics systems could interact with each other, or better yet, both be implemented in the same vehicle/prop, issues with unrealistically deforming props in excessively hard collisions, and the annoyingly bouncy cannon, could be mitigated. Also, hardbody physics would be less resource intensive, so it could be used for destructible map elements. This could be a major addition to increase immersion and make the maps feel more dynamic. Again, this would likely require a major rework to the physics engine for a relatively small problem, and I have little expectation that this can actually be implemented any time soon, but I thought I'd put my opinion out there.
