Running the `boxes2` example twice gives two different results

[aleksa2808]

As stated in the title, running the boxes2 example back-to-back gives two visibly different results. Running the 2D CCD example from the rapier repo gives the same result every time. Is it possible there is a bug in the bevy plugin layer that's causing the difference?

First run:

Second run:

[sebcrozet]

That’s likely related to #233

[rparrett]

FWIW, I can reproduce the behavior on an M1 mac.

[aleksa2808]

That’s likely related to #233

Perhaps, although in the current state of that pull request I'm still getting differing results with the boxes2 example.

[aleksa2808]

As stated in the Bevy discord's #physics channel, when I add the RapierConfiguration resource with TimestepMode::Fixed (tried with dt = 1.0 / 60.0 and substeps = 1 values) each run is now the same.

The question now is: was I wrong in assuming that two boxes2 example invocations with the default plugin configuration would provide the same result? If so, then there is no bug present and I can close this issue.