Multi-shot acquisition for diffusion MRI is challenging due to shot-to-shot phase variations caused by motion. Multiplexed sensitivity-encoding (MUSE) tackles this problem by extracting phase estimates from reconstruction of individual shot images. In this approach, the feasible number of shots is limited by increasing g-factor noise penalty. Against this background, the present work studies the feasibility of highly segmented MUSE with spiral acquisition, which offers particularly benign g-factor behavior. To stabilize reconstruction, we explore the utility of moving from separate estimation of phase offsets to joint optimization of phase biases and image content.