The query optimization team has invested many engineer hours writing microbenchmarks, which are very useful for spot testing potential performance improvements. However, change points are not currently triaged so their value diminishes over time. 

The idea with this ticket is to identify a subset of QO-specific microbenchmarks which are deemed valuable and attempt to stabilize the tracked metrics (potentially by switching to instruction count instead of latency). This would be a requirement to the eventual goal of tracking these benchmarks as "performance correctness" tests which would trigger a BF on a regression similar to the existing SEP ping benchmark.