First of all, as a software engineer I’m — well, “impressed” is the wrong word because I remember how efficient software used to be in the '90s — I’m “satisfied” with how well Lemmy instances are scaling. Even the largest instances are running on single, fairly-small servers.

Keep in mind that this is all alpha software and not only likely very unoptimized but also pretty buggy, so the surprisingly few problems there have been are more likely due to that than to real issues of scale.

Second, and more importantly, remember that having really big instances is “doing it wrong” to begin with. The intended design of Lemmy (and Fediverse services in general) is to have a whole lot of small instances, not a few big ones.