Quick answer: Physics bugs are notoriously hard to reproduce. A test plan that explicitly varies framerate, time scale, and entity count surfaces 80% of them.

Physics bugs hide in the parameters. Vary them deliberately.

Vary framerate

Test at 30, 60, 120, 144 fps. Framerate-dependent bugs surface here.

Vary time scale

Slow-mo at 0.1; fast at 5x. Some bugs only appear at extremes.

Stress entity count

1, 10, 100, 1000 physics bodies. Performance and correctness both visible.

Vary input timing

Hold inputs steady; tap rapidly; combo presses. Input-timing-sensitive bugs surface.

“Physics is parameter-sensitive. The parameters are your test plan.”

Build the parameter-variation test plan once. Each release runs through it; the same bug class doesn't slip twice.

Related reading