[ODE] Possible solution to unstable rotating bodies problem

[gl]

> The correction velocity in problematic cases usually goes upwards and is
> automatically cancelled by gravity. Since correction is always done over 
> few
> steps, and if there is residual penetration as well, then the object will
> not bounce at all (unless you explicitely want bouncing, and add that). 
> When
> objects are penetrating sideways, the correctional velocity will still be
> there, but 1m/s is very low and is much better than the usual velocities
> caused by teleportation, which can be on the order of 100m/s and up, YMMV.
> Overall, this thing works better than it looks at first.

Alan, I agree it's a better solution than none at all, and it probably works 
well for most cases - but again (like the other issue being talked about) it 
breaks if the sim is set in space, and I think to just avoid the issue 
altogether and make error correction 'do the right thing' is a good idea.

If I get time, I'll try to implent it and see what happens.  But if someone 
else wants to have a crack at it now, even better : ).
--
gl