Instability in a setup similar to the vehicle_sample after some time

[maxmus1983]

We do a tank simulation and are evaluating box3d as a physics engine. I implemented the tank as a box with 14 "wheels" as seen in the screenshots. Our terrains are fairly big (dozens of km), but that does not seem to be the problem (we use BOX3D_DOUBLE_PRECISION) and the problem also happens when we start at low positions (< 50m in all dimensions).

Our setup idea is to make the suspension like in the vehicle sample just calculate the "up" force and then calculate the "forward" and "hull rotational" force by our engine and apply it.

In the beginning the tank is very stable and is driving forward as it should, but after some time it begins to shake strongly and then flies into the air. The only force I am applying is the "forward" force of 0.2g (and of course gravity).

We also stream terrain, but that seems to work ok in the recording viewer (the y-value is always the same in the rec sample for simplicity)

b3recording.zip

If the screenshots and recording are not enough i could create a simple sample code like in the vehicles sample, with a constant forward force. That also shows this behavior.