Problem: when there are many curves, the mapping from X/Y/Z/W to colour is inconsistent. For example, the same hue can be used for X in one bone and Y in another.
Proposal: No concrete proposal yet, apart from the desire to have consistent X/Y/Z/W colours. The colour generator code could be altered to always produce the same hue for the same RNA array_index, for example, or use array_index/array_length.