As you'd expect with a 25+ year code base there are aspects of the MySQL codebase that are unglamorous. These dark corners contain implementations that introduce behaviour that users depend upon, that changing will introduce storage or replication compatibility issues, or just don't attract the resources to sufficiently rewrite. These areas of code can lead to performance degradation by having code that doesn't utilize hardware in an efficient way, or is written in a way that limits scalability. Through these examples, I'll cover techniques of constructing microbenchmarks using sysbench, using perf, and experimentation with MySQL code. Using these techniques, many aspects have already been fixed and I hope you'll be the next person to write a bug report, potentially including a patch, that enables these corners to become a little bit brighter.