Exactly. In fact I'd go so far as to say this is an equally Important Thing to the Important Thing in the article. If your queue is just a simple FIFO / LIFO / whatever to buffer surges in demand until you get to them, then you absolutely get the "queue suddenly goes to infinity as you hit a critical point near maximum throughput" thing happening.

If, however, you use the queue as a way to process queries more efficiently then that's a different story.