This would be comparable to lemmy.ml accepting/rejecting accounts based on whether the account email is hosted in "gmail.com", "hotmail.com" or whichever other third party service.
That's true if nothing is expected from the user-instance side. But for example a community-instance may expect from the federated user-instances that they ban user who are repetedly reported for bad behaviour.
This will be all the more true when Lemmy will federate with other fediverse platforms, such as Mastodon. There, there is a public user to user interaction and so it may be desirable to block the instances that allows undesired behaviours, because most users from there will make no difference and act the same way on Lemmy communities.
The user-instances in my example also do not serve content from other instances, so they are also very unlikely to see a need to block anyone.
They serve content from the community instances to their users, whether they host it or not.
If you really want those instances to disappear altogether and the user accessing communities from client, we lose the feature of user-instances blocking users for all their members, and the members will have to block e.g. all nazis one by one.
Remember the reason we talked about this: if you don't duplicate communities then the fact that the amount of popular active topics is limited can lead to huge centralized nodes forming around the active communities.
They can form, true, but they can also stop growing at some point, as a result of either their own will, the new communities being created elsewhere or cpmmunities migrating elsewhere. In fact, shared communities also don't prevent these instances to become bigger and bigger.
Those 2 users can't be active in an infinite amount of topics.
Of course I didn't mean that litterally an infinity of communities will exist some day, just that there are way more topics than what is currently on lemmy.ml so there is a lot of room for other instances to grow.
Also, the limit of topic that they can discuss simultaneously is not the same as the global one, considering that new communities (dis)appear everyday. And sometimes one will create a new community instead of one that is dying, e.g. setting new rules that they think will improve it. Maybe creating a community on a new instance with different CoC.
I think I pretty much inderstand you now. The point of my 2-item list was to draw a line between "communities submitted to the meta community" (which I also referred to as post in that part), and the ones who are actually part of the meta. I was thinking "cutoff" rather than "weight for appearance in the feed", but the latter is also interesting !