Question: what can differential dataflow be used for?

This is a fascinating project. I've been studying differential dataflow and specifically the core operators like iterate and arrange. I've also been trying to map out how it juxtaposes with reactive programming (a la MobX).

Can differential dataflow be used to implement, say:

• a programming language compiler with incremental compilation (for fast as-you-type intelligence, fast builds, etc.), but written "naively" as a non-incremental multi-pass compiler?
• the core React reconciliation algorithm (used to be referred to as "virtual DOM"), but written "naively" as if it were generating the tree just once?

Most of the examples I can find on differential dataflow show how to perform some generic data crunching with very simple input and output structures, but I'm asking more about transforming some rich data structures into other rich data structures.

Start with the differential dataflow examples and the iterate and arrange operators named in the issue; compare their simple input/output models with the compiler and React reconciliation use cases. Done would be a documented answer, ideally with a concrete rich-structure example or explicit limits.