In case you are interested in the full story:

It does not matter much to the user, but it does matter a little. Which is also why it gets hard to find a good solution.

The communities are the same, and posts are synchronized. However, each server (also known as instance, to make it more complicated) is responsible for its own moderation.

Say for example that it turns out quokk.au is run by the absolute worst people, and that they start encouraging vile content and hate speech on their platform. Or simply that they don't moderate, and a lot of bad actors start signing up there and posting spam. Other servers might say in response that they don't want to keep seeing content from quokk.au, as bad content keeps coming from there. Of course that would affect the users of that instance.

To a degree, these problems can result from size. Lemmy.world is the biggest instance of Lemmy, and while it's generally decently moderated they have of course had greater challenges making sure their user base behaves decently as a result. So we want users to spread out so that each instance can have human moderators overseeing a manageable amount of human users.

This means that your instance can matter: if it is particularly poorly managed, you might be cut off from parts of the network as a result. And on the flip side, if you join an instance with stricter moderation, you might see less questionable content from elsewhere. Typical divides is whether Russia propaganda networks like lemmygrad.ml are blocked form joining the network, and whether porn is allowed.

The other big reason we want people to go to different instances is to keep the network healthy. If (almost) all of us are on the same site, that gets expensive to run, and leaves too much responsibility and power in the hands of the people running that site. If something happens to it it's a huge problem, if moderation fails it's a huge problem, and it's just generally a weakness. We want people to spread out to a bunch of different sites so that the network itself is robust.

Over on Mastodon, an anarchist instance hosted in America had the police raid their physical location and steal all the data off their server. Obviously we're also less vulnerable to these types of attacks if we spread out, though more often the problems are more boring and ordinary.

So in effect, when people ask to sign up at piefed.social, the good people at piefed.social wants people to join the network, but preferably the load should be put on some other server in that network in order to distribute the risk and workload, and strengthening the network as such. This is not primarily for the immediate benefit of the individual users—though it will materialize as such though more human moderation and admins that have the capacity to actually care about them—but for the benefit of the network.

So that's why we struggle to find a good way to ask people to pick a server. The choice is not super important for the user—the recommended instances are all decent, and the little information provided should be enough to make an informed enough decision—but it's important to the network that some users make a choice.

If you have any recommendations how to make this easier while distributing users around the different servers, feedback and ideas are super welcome.

This would also require many instances to set out in detail their location, focus etc and have a category system that recognises that.

This is already the case. Each instance sets their own trusted instances. There are no instances trusted by default. In /admin/instances if you click edit in an instance's row, then you can check the box that it is a trusted instance. Also, these are one-sided relationships. Instance A can mark Instance B as trusted, but the converse is not required.