Can't find the post right now but sunaurus said that the main difference is that lemm.ee is the only instance using a horizontal setup, that is, there's multiple lemmy servers running on multiple servers behind a load balancer, all sharing a database (postgres itself clusters very well). The code isn't actually made for that so it's all rather custom and possibly specific to his hoster.

Less technical: sunaurus happens to be a beast of a devop and prolific contributor to the lemmy codebase. As such lemm.ee quite often runs code that's ahead of the release schedule, addressing stuff that he stumbled across while wearing his sysop hat.