- There aren't really any flashy results (the whole thing could be sold as restatements of "if the average processing rate is similar to the arrival rate, any variance in arrivals will lead to long queues".
- As a corollary, most queues encountered in practice can also be dealt with using very simple techniques or making the queue negligible. Neither of which require the study of queue theory.
- The quick recommendations are really boring (if you want the queue to get shorter, you need to process faster or add more servers).
But studying queue theory handy nonetheless because it turns out that queues are, in practical systems, about as common as list data structures or associative maps in programming. They are everywhere. Every time a stream meets a buffer, in fact. Being able to see a situation and reducing a lot of the noise to (M/M/n, lambda = 0.2, mu = 0.4) can free up a lot of thinking horsepower for more interesting problems. Then there is no need to try to reason about queue lengths vs serving times vs variance vs how those change with the addition of servers. An expert understands that a lot of results flow from a few simple variables, and doesn't have to remember the details because they are just symptoms of a few key observations.
So, in a sense, the reward for knowing a lot of queue theory is not having to think very much about queues.