SummaryAnimals usually are not willing to perform at levels, or for lengths of time, of which they should be maximally capable. In stating this, exercise performance and inferred capacity are gauged with respect to body size and the duration of particular levels of energy expenditure. In such relative terms, the long-term metabolic ceiling of ca. 7 times basal metabolic rate in challenged but energy-balanced individuals may be real and general, because greater performance over long periods requires larger metabolic machinery that is ever more expensive to maintain. Avian marathon migrants relying on stored fuel (and therefore not in energy balance) that work for 9 consecutive days at levels of 9-10 times basal metabolic rate are exceptional performers in terms of the 'relative expenditure' on 'duration of a particular activity' curve nevertheless. Here I argue that metabolic ceilings in all situations (energy balanced or not) have their origin in the fitness costs of high performance levels due to subsequently reduced survival, which then precludes the possibility of future reproduction. The limits to performance should therefore be studied relative to ecological context (which includes aspects such as pathogen pressure and risk of overheating), which determines the severity of the survival punishment of over-exertion. I conclude that many dimensions of ecology have determined at which performance levels (accounting for time) individual animals, including human athletes, begin to show physiological restraint. Using modern molecular techniques to assay wear and tear, in combination with manipulated work levels in different ecological contexts, might enable experimental verification of these ideas.