Yeah, if you reason about each task as a thread, then all the same problems with sharing data between threads applies to sharing data between tasks, regardless

Yeah, if you reason about each task as a thread, then all the same problems with sharing data between threads applies to sharing data between tasks, regardless of single threaded runtime or not. The only real difference is you can avoid atomics. Generally, I advocate for avoiding sharing any non-trivial state which would necessitate holding a "lock" across an await point (there are rust lints to help you here if you use RefCell). This usually requires an actor pattern to explicitly queue up work that requires that state into its own self contained task rather than reach for an async mutex.