The successful completion of complex tasks like hanging a picture or laparoscopic surgery requires coordinated motion of more than two limbs. User-controlled supernumerary robotic limbs (SL) have been proposed to bypass the need for coordination with a partner in such tasks. However, neither the capability to control multiple limbs alone relative to collaborative control with partners, nor how that capability varies across different tasks, is well understood. In this work, we present an investigation of tasks requiring three-hands where the foot was used as an additional source of motor commands. We considered: (1) how does simultaneous control of three hands compare to a cooperating dyad; (2) how this relative performance was altered by the existence of constraints emanating from real or virtual physical connections (mechanical constraints) or from cognitive limits (cognitive constraints). It was found that a cooperating dyad outperformed a single user in all scenarios in terms of task score, path efficiency and motion smoothness. However, while the participants were able to reach more targets with increasing mechanical constraints/decreasing number of simultaneous goals, the relative difference in performance between a dyad and a participant performing trimanual activities decreased, suggesting further potential for SLs in this class of scenario.