In all honesty, this is probably most useful for things like, self evaluating trees. Based on some input of events, create a tree of how things should evaluate and execute upon that.

Imagine creating a flow chart, and you wish to see how it evaluates later on in a test scenario. Heck, it's implementable deterministic finite automata. If you somehow put a step function to control in there and have enough memory, you could do some non-deterministic finite automata.

By and large, from an everyday, custom, programming life, I agree there's no easy usage, but for people who make software engines, this is probably more useful.