...while avoiding to flood the result queue (this is why this code has been
removed originally)

This is possible since the manager coroutine may no longer outlive the executor
instance.

While this is useful to shut down a management thread when the executor gets
garbage collected, this techinque is not suited for a management coroutine.
For it to work the executor must be collected while the event loop is still
running.  In practice this will be often not the case and the shutdown will not
work as intended.

Instead, it is recommended to always use the async context manager that ensures
correct shutdown.  The i_will_remember_to_shutdown=False argument was added to
ProcessPoolExecutor.__init__ to help detect mistakes.