If you've seen other posts that I have made you might've noticed that I'm interested in this exact problem and I have been super focused on research that shows having efficient information networks (i.e. centralized networks, a network where only one or a few voices matter, or fully connected networks, a network where everyone can see everyone else opinions) can lead to much lower collective intelligence for the group and having inefficient information networks (networks that have fewer connections, maybe 4-6, and everyone has an equal amount of connections) can lead to a group being able to solve more complex problems.
So in relation to what you pointed out big online communities actually might be making our collective intelligence weaker even though it makes us more connected.
I think that the individual's ability to process information does play a role in how many connections that individual should have but the more important role in having fewer connections is to provide protection from social influence which can hinder the creativity process and help stabilize adoption.
So for example, if I had 50 connections and 4 people adopted a new behavior / shared new information I would still be influenced to not take up the behavior because so many of my other connections aren't taking part and it could lead to negative feedback from my other connections, but if I had 6 connections instead that behavior/information would be much more appealing allowing for newer ideas/behaviors to spread in a much more stable way.
Similarly with creativity. If you have a lot of connections that are giving you answers to everything you could think of (and they are decent answers) then there is less of a need to find creative solutions to those problems meaning that new ideas are less likely to be thought of or proposed. Alternatively being surrounded by that much external information siloes you to think about finding a new solution within the things that have already worked.
This is something that they have sort of studied but not in the way you have suggested. They have allowed individuals to change their social connections over time and have noticed that the connections become more centralized and/or the connections are to people who are like them in relevant ways (This point isn't in this paper but it is in some related research). It would be interesting to see what would happen if they actually optimized the network over time to make everyone smarter.
Yes, I find it very interesting too.