Just came across a #pico8 oddity (not quite a bug).

The value of stat(1) is based on the number of cycles since the start of the frame, so if the cycle count overflows, stat(1) becomes negative.

I did expect an overflow, to be sure, but it took me a while to figure out *where*.