I would love to add a good RRB implementation to the persistent benchmarks at [1] to get a state-of-the-art comparison between RRB and BST in a persistent context. Duration, of course, but also number of bytes copied etc.
I always wanted a comparison to ropes. Every time I see ropes mentioned I always think "why not use RRB trees?". It seems like less housekeeping, but with all the benefits.