I can’t say I have, but I basically only browse from my instance, so I may very well be missing out on things I could see on the instance the community is based on. I have noticed that .world and .ml are usually very slow to return results for my searches (as well as some other large instances like sh.it-just.works), and I have had trouble getting them to acknowledge my subscription (retrying a few times over a few hours or days works). I think this is basically because they are still overloaded, at least periodically. If .ml is slow to federate because it has lots of federation work to do and .world is failing to accept requests due to load that sounds like a recipe for sending of posts, comments, votes, etc. to have to retry over a period of minutes or hours, if at all.

If you are running your own instance it would marginally increase the federation load on e.g. .ml in the example above, but since the server you run isn’t overloaded you would most likely see things on your instance after .ml’s first attempt to send you the post/comment/vote/whatever. The ideal would be lots of medium sized instances that can handle the load so that there isn’t too much federation work to do (having only one user per instance would mean servers need to federate to thousands or even millions of servers, which would be a lot of work and bandwidth), but at the same time no single server would be too overload to handle the incoming messages either.