How would this be different in a different concurrency model?
That was actually the point. You have the same space considerations. Being coroutine instead of native thread is just a small constant factor... if this factor is important then you are almost certainly doing it wrong.