mass, force and tire load

[NaN]

Duh, should have read your post more carefully. The problem you observed is most probably due to the fact that the GAME has a separate fixed time step logic. Whereas the physics engine has got its own. In your case the physics engine would be getting timesteps shorter than its fixed timestep, would drop simulation frames. It is interesting that it worked at all. Ideally one would feed the game logic as prestep, poststep callbacks to the physics sim, use one fixed step implementation.