Human movement is characterized by its variability: the same task is never performed twice in exactly the same way. This variability is believed to play a functional role in movement performance and adaptability, as well as in preventing musculoskeletal damage. This article focuses on the time-evolution of movement variability throughout a repetitive pointing task until exhaustion. The kinematics of 13 subjects performing the pointing task is analyzed. Principal Component Analysis of joint angles identifies joint coordinations for each pointing cycle, and cycle-by-cycle comparison highlights movement variability. Non-supervised clustering reveals that subjects adopt successive coordination patterns at an intra-individual level. Inter-individual variability is characterized by the number and type of such patterns: from 3 to 5 patterns, mobilizing the trunk, the shoulder and the upper limbs differently. Movement variability exists even in a seemingly basic and constrained task. It appears in the very early stages of fatigue onset, and may correspond to adaptative coordination responses throughout task performance. This observation should encourage workstation designers to better account for movement variability in order to preserve operators' health and safety.