code; the question is whether we guarantee that the completion has
*already run* or not - we don't want to do this on the processor, as
anything that involves changing Task status can lead to thread theft, so
it is done via the custom thread pool; the fix here is simply to check
that the task *becomes* cancelled (in a timely fashion), rather than that
the task *is* cancelled; anyone using "await" or ".Wait()" ".Result" will
observe that *anyway*; also tidied up some efficiency code to reduce exception
allocations in this scenario