[ODE] Torque instability

[nlin@nlin.net]

Hello,

In attempting to reproduce Ruud's problem (which I was unable to), I was
experimenting with applying relative torques to bodies floating with no
gravity or forces. If I apply a relative torque of 50 (with dBodyAddRelTorque
(50,0,0)) to a box with mass set via dMassSetBox(&m,5.000000,0.218034,0.239283,0.509767),
the box starts to spin very rapidly and soon thereafter disappears, with an
orientation matrix consisting of all "nan". Does this mean my timestep is
too large for stability, or could there be another reason for this behavior?

-Norman