High-intensity exercise tolerance is described by the hyperbolic power-tolerable duration (P-tLIM) relationship. Tolerable duration is proposed to be defined by the expenditure of a fixed amount of work, reflected in the curvature constant parameter (W′) of the hyperbola. However, it is unclear whether W′ depletion at task failure represents the attainment of the physiological limits for producing the required task power in all high-intensity tests, or whether a ‘reserve’ in the capacity to generate power exists. Therefore, we investigated (1) whether the physiological limits were attained at task failure in four different high-intensity tests and (2) the relationship between fatigue and W′ depletion, in tests with different W′ depletion rates. In nine healthy participants, maximal voluntary isokinetic cycling power (PISO) at 80 rpm was measured before and immediately at task failure of four constant-power cycling tests of differing power and tolerable duration. The constant-power tests were used to characterise the P-tLIM relationship, while the reduction in PISO from baseline provided a measure of fatigue and identified the presence of any reserve (task failure PISO vs. task power). Subsequently, the power predicted to result in task failure at 6 min (WR6) and 12 min (WR12) was estimated from the P-tLIM relationship. During completion of these WR6 and WR12 tests we measured PISO at baseline, 1/3rd and 2/3rd of expected W′ depletion, and immediately at task failure. Baseline PISO was not different across the six constant-power tests (mean ± SD: 837 ± 108 W; p > 0.05). Similarly, VO2peak was not different across all six constant-power tests, despite the differences in task power (4.23 ± 0.52 L/min; p > 0.05). Fatigue developed to the extent that task failure PISO was not greater than task power (i.e. no reserve) in the two higher-power, shorter duration tests (338 ± 53 vs. 325 ± 35 W and 348 ± 57 vs. 305 ± 34 W; both p > 0.05). A reserve was present, with fatigue insufficient to limit the task, in the two lower-power, longer duration tests (356 ± 85 vs. 282 ± 34 W and 420 ± 89 vs. 266 ± 37 W; both p < 0.05). During WR6 vs. WR12, PISO was not different at 1/3rd (623 ± 119 vs. 606 ± 94 W; p > 0.05) and 2/3rd (485 ± 95 vs. 479 ± 111 W; p > 0.05) W′ depletion. However, task failure PISO in WR6 was less than WR12 (360 ± 75 vs. 413 ± 99; p < 0.05). Reducing the task power and W′ utilisation rate, thus increasing the exercise duration, led to task failure at VO2peak with a power reserve (task failure PISO greater than task power). This suggests some tests are terminated at VO2peak with a reserve in the capacity to generate power, and mechanisms other than muscle fatigue are contributory. The time course of the fatigue-induced fall in PISO during WR6 vs. WR12 suggests these alternative mechanisms of exercise limitation become more prevalent late in the exercise test.