Twisted and coiled polymer muscles (TCPMs) are a type of artificial muscle with a remarkable power-to-weight ratio. However, actuation dynamics are slow compared to other artificial muscles. This work aims to improve dynamic performance by incorporating redundancy. Specifically, this work examines if TCPM bundles of heterogeneous geometries containing high-force low-bandwidth actuators and low-force high-bandwidth actuators have a substantially better tracking performance than that of bundles of homogeneous geometries. First, a white-box model was created to simulate TCPM dynamics as a function of geometric parameters. The model revealed fiber diameter is the only geometric parameter that represents a trade-off between TCPM bandwidth and maximum realizable force for isometric force tracking. Next, an optimum feedforward controller was designed to distribute the reference among redundant actuators. Finally, a brute-force optimization was conducted to find the optimum configurations of heterogeneous and homogeneous TCPM bundles and the associated tracking performances. Optimal homogeneous configurations outperformed all heterogeneous configurations irrespective of number of TCPMs in parallel or reference signal. For unidirectional configurations, a nontrivial fiber diameter optimizes tracking performance. For antagonistic configurations, tracking performance improves monotonically with increasing fiber diameter.