Are the registry procedures causing the problem? Since most of my communities are on lemmy.ml, wouldn’t I still be using lemmy.lm’s bandwidth? Or is it more crucial to ensure that the communities are distributed more evenly across different instances?
The keyword is “cache”. Lemmy doesn’t guarantee the cache, so if an instance X would go down permanently, you would eventually lose this data.
My presumption was that with over 200 instances, maybe over 1000 in the near future, all the old data should be collectable in theory without loss. But you’re right, there’s no guarantee of this.