Could you please elaborate on the intricacies involved in the bitonic merge process? I'm particularly interested in understanding the level of complexity it presents, especially in terms of algorithmic efficiency and computational demands. How does it compare to other sorting or merging techniques in terms of its complexity? Additionally, could you highlight any specific challenges or considerations that arise when implementing bitonic merge in practice?