Running biomechanics and its evolution that occurs over intensive trials are widely studied, but few studies have focused on the reproducibility of stride evolution in these runs. The purpose of this investigation was to assess the reproducibility of changes in eight biomechanical variables during exhaustive runs, using three-dimensional analysis. Ten male athletes (age: 23 ± 4 years; maximal oxygen uptake: 57.5 ± 4.4 ml02·min-1·kg-1; maximal aerobic speed: 19.3 ± 0.8 km·h-1) performed a maximal treadmill test. Between 3 to 10 days later, they started a series of three time-to-exhaustion trials at 90% of the individual maximal aerobic speed, seven days apart. During these trials eight biomechanical variables were recorded over a 20-s period every 4 min until exhaustion. The evolution of a variable over a trial was represented as the slope of the linear regression of these variables over time. Reproducibility was assessed with intraclass correlation coefficients and variability was quantified as standard error of measurement. Changes in five variables (swing duration, stride frequency, step length, centre of gravity vertical and lateral amplitude) showed moderate to good reproducibility (0.48 ≤ ICC ≤ 0.72), while changes in stance duration, reactivity and foot orientation showed poor reproducibility (-0.71 ≤ ICC ≤ 0.04). Fatigue-induced changes in stride biomechanics do not follow a reproducible course across the board; however, several variables do show satisfactory stability: swing duration, stride frequency, step length and centre of gravity shift.