When a waiting-time test sees dissipation that is not there. An Evidence Press briefing. Tiny machines, such as molecular motors in cells or beads held in laser traps, constantly exchange energy with their surroundings. Physicists measure how far such a system is from equilibrium by its entropy production: the rate at which it turns energy into heat. Usually only a few of the system's jumps can be seen, so researchers try to infer entropy production from the visible jumps and the waiting times between them. In 2024, Maier, Degünther, van der Meer and Seifert proposed a convenient statistic for systems driven by a repeating rhythm. It needs only the ordinary experiment, not a second experiment with the rhythm run backwards. After testing more than a hundred thousand random models, they conjectured that it never overstates the true value. This release shows that the conjecture is false. In a three-state model, two hidden energy barriers are raised and lowered out of step. The system stays in equilibrium at every instant, so it dissipates nothing. Yet the visible waiting times look different in the two directions, because the drive makes the hidden barriers open in a fixed order. The statistic reads this as dissipation. The effect in this example is very small. But in the class of models the conjecture's authors tested, weak driving is always safe, while slow, strong driving overstates dissipation, by at least thirty-one per cent in one case certified with interval arithmetic, and by about a factor of two in numerical examples. Estimators that use the reversed rhythm remain valid. The practical lesson is not to report this forward-only statistic as a lower bound on dissipation. These are mathematical results about model systems, not experiments. This is an anonymous, AI-assisted, unrefereed candidate, not peer reviewed or independently reproduced. Published on 21 October 2026, its paper and evidence are linked from the Evidence Press release page. This briefing uses an AI-generated voice and is not additional scientific evidence.