What I mean is, why not just a bi-directional mirror, with several servers in the larger users goes? Can the federation API be leveraged to just do that to address the dramatic load increase? Folks firing up Lemmy servers all over that only have a tiny portion of unique content are likely to provide poor experiences and stress the federation subsystem on the larger lemmy.world and lemmy.ml sites.
I think the idea of a massive instance is discouraged somewhat
Are you wondering if one instance can be spread across multiple machines? I don’t think that’s possible and seems unnecessary. I say that because spreading users onto more instances creates less load on the larger instances, not more. Edit: I guess you can scale instances across machines!
Instances can be scaled across several machines. Here they’re using multiple containers as their resource usage isn’t high enough to require multiple machines yet, but it proves that it can be done.
A lot of the value isn’t technical, it’s social. Each instance can set their own rules for acceptable conduct and what sort of content they want or don’t want. That’s one of the most valuable parts of decentralization, an instance like Beehaw can try to be an open and inclusive space and thus have a longer list of rules, while another instance can be more permissive and allow NSFW and more offensive speech. And thus the two can coexist in the same network with the same namespace.
