Having glanced at the code and taken my own instance down a few times cleaning up a surprisingly small number of automated posts, it's definitely the combination of some design choices in the code and the scale of lemmy.world. Keeping an instance up that has so many posts and communities has been difficult on my instance, and I'm basically the only user. I can imagine with the scale and lemmy.world load and publicity, it's nearly impossible until some improvements to the data layer are made.

For one example, purging a community with 1k posts and 30k comments (I was messing with a bot) took my instance down for 2 hours with the postgres database pegged at a full core minimum. And then it took down my instance. And then I restarted the database but presumably this was done in a transaction so no progress had been made.

I'm personally impressed with the amount of uptime lemmy.world has managed. And I'm also impressed with lemmy overall, but it's pretty clear there has been some rapid growth that, as it usually does, exposed some of the limits of the design and requires some improvements for the current scale.