Currently if an operation is load shed it means that all the work that have been done so far will be thrown out, which in an overload scenario is complete waste of resources and effort.

The purpose of this ticket is to implement logic that exempts re-entering operations from this limit. Specifically, operations that have already consumed resources and yielded execution will bypass the queue depth check when attempting to re-enqueue.