From the documentation, it doesn't seem clear that there's any guarantee that a particular queued item will ever eventually be dequeued (in a program that runs forever).

Consider the case where each producer thread queues N items, and then waits until at least one of its N items is dequeued before immediately topping back up; while the consuming thread dequeues at a slower rate than the producers are able to produce. Maybe no item from producer number 1 ever gets dequeued? Or did I miss something in the documentation?