Simulation is not deterministic if many explosions

[Wuzzy2]

If a lot of dynamite explodes or a lot of cola bottles spill their cola everywhere, the simulation is no longer deterministic. If you repeat the exact same situation, you might get different results.

How to test

I have added a test level levels/test/test-dynamite-determinism.xml in the master branch to test this. Just start the simulation at normal, fast or ultra fast speed and look where the tennis ball will drop. Repeat the simulation multiple times.

EXPECTED: The tennis ball always ends up in the same spot
ACTUAL: The tennis ball ends up in different spot

If you only use one dynamite, or the level is generally very simple, it seems to be deterministic.

Version

123d276 in master