Thirty-eight of forty-one captures show the same stall at almost identical intervals along a fixed traversal route. The average frame rate stays above target throughout, which is why nobody writing from feel has identified the cause.
Button-to-photon latency measured externally at 1,000 frames per second, 2,400 samples per title, three input devices. The spread between fastest and slowest is 41 milliseconds.
Turn resolution timed across fourteen late-game saves with identical action sequences. Cost scales with tracked entity count at roughly the square, and the curve becomes noticeable long before the interface warns you.
Twenty-two captures across four titles show that the expensive moment is not rain but the two seconds in which rain begins, where shader compilation and particle spawning coincide.
One hundred and eighty timed runs across fifteen titles, comparing first launch after reboot against subsequent launches. The difference averages 3.4 times, and almost every published figure describes the easier case.
Thirty-six recordings of a fixed sound source behind varying geometry, timed and compared against the expected attenuation. Two of eleven material types behave incorrectly, both of them glass.
FrametimeDesk