Edit: searched the forums and found that it's caused by illegal math operations like div 0 etc. Also found a bugfix in 0.99 which fixed a problem with anglediff() sometimes returning -1.#IND with 2 identical parameters due to a float rounding error. This sounds exactly like my bug, since i'm using long floats with anglediff every frame, with a high possibility of identical angles. I'm geussing the bug is actually still there.
Using Classic r1.2
Also, can you check for -1.#IND in expressions/conditions as a string? I'm guessing no, but it could be useful.