Task continuation on a different thread

[nelsc]

We should carefully consider guaranteeing that the thread that executes the code following a task block will be the same one that entered the task block in the first place. This would effectively guarantee that an arbitrary function call will not return on a different thread, even if the function involves parallel execution, and would drastically limit the surprises that come with parent-stealing.

(On this topic, I recorded this sentence fragment, without fully understanding it: "Processors stealing threads instead of threads stealing work")