Yep, agree with those crazy ideas. They are worth exploring!
I think the granularity is also important here. We have been taking about tasks in the above example because both tasks only do a single thing and then return. For the full range of testing we might want to work on the “job” granularity I.e. until the next suspension point. Executor hopping is also another interesting edge to consider and we would need to be able control all executors in a test even those of actors