So instead of requiring the callee to be async or not you have to make sure your caller is executing in the right effect handler context? Sounds like we replaced the coloring problem with another one? I can see how it benefits the language designer because you could implement try catch and async with this but how does it benefit the user?