I’d imagine the idea was to keep the design flexible. 1.

Job

is lightweight and it’s more like the representation of the state of the coroutine.

Deferred

itself is actually a

Job

, but with more capability. It holds the result of the execution 2. The behavior of

async

and

launch

is slightly different in that

launch

will propagate any exceptions thrown, whereas

async

saves that result/exception in the

Deferred

. This also makes sense given the description in (1) above. This is only my understanding of it. Perhaps someone will put up a more detailed description of the design decisions.