asko wrote: > > It doesn't seem to make any difference wether I use the default > approximation or the friction pyramid one, the problem exists as long as > there is enough friction to slow down the box. What's the smallest piece of code that can reproduce this behaviour? - Martin