> Would you try python_startup and python_startup_nosite benchmark

That seems entirely unnecessary.  If adding __reversed__ has any effect on the rest of the build, that is pure random noise that can be ignored. It is normal to get small improvements or detriments that vary from compiler to compiler, os to os, etc.  Our normal practice is to ignore those and not optimize for random or hard-to-control compiler idiosyncrasies which appear and disappear as we add methods or as compilers get updated.