I would love to see an in-depth analysis why Go's GC behaves this way for this specific benchmark in the first place. Just playing around with different values

I would love to see an in-depth analysis why Go's GC behaves this way for this specific benchmark in the first place. Just playing around with different values for `GOGC` seems like guess-work and could just paper over the underlying real issue. I know this probably requires to read the GC's source code, but that would've been certainly been very educational.