About 10 years ago I was mentored by a guy who was an utter wizard at queuing theory, and who bugfixed a whole bunch of nasty issues in cellular telecoms hardwa

About 10 years ago I was mentored by a guy who was an utter wizard at queuing theory, and who bugfixed a whole bunch of nasty issues in cellular telecoms hardware through his understanding of how queuing theory impacted code execution. TL:DR - queue behaviour gets nonlinear as you approach the theoretical max load. If you are running your processors at a high load, even a small change in code throughput makes a huge difference to real world behaviour.