> I’ll try and write a followup with benchmarks.

That would definitely keep the story from being all-hat-and-no-cattle. I can't recall reading something with so many alternate versions of how to implement something but with zero benchmarks.