Improve a QA benchmark and regression setup

A role in quality assurance for hardware and software: discuss how you run benchmark and regression testing, and how you would improve it. Questions reported: how you check that benchmark results are valid and judge performance; how you cut the false failure rate when regression results are not exactly the same from run to run (for example, accepting any of a known set of correct outputs); how you classify test outcomes; how you reduce the shared compute cluster time the suites use; and how you shorten each benchmark run.