[Groxx]

Though my use of stranger data structures is relatively rare in my line of work, I enjoy learning them because they keep revealing new ways to approach problems. It's less "I use CRDTs" and more "CRDTs have changed what I think is feasible", similarly for stuff like bloom filters (no-false-negatives can be very powerful) and locality-sensitive hashing (don't search for all X every time, pre-compute a simpler version so you only check promising items). Concurrent data structures have also been pretty fruitful, since they're also often "how can we guarantee X in our distributed system, and what primitives do we need" strategies / solutions.

(obviously some of this is "I learned a tangential thing while learning about X", but that's kinda the point. you may never use a rope, but you might be introduced to a new idea in the process of learning about it)