Exactly.
Stake (de)centralization is not a function of N number of nodes. It is a distribution within the N set. Increasing N would actually actually increase the amount of effort/work to decentralize stake, as it now has to be evenly distributed among more validators. It seems that the proposers have this logic inverted. As is, it’s akin to saying right now we can’t evenly feed 100 ppl, but if we try to evenly feed 200 ppl we should be able to get it done.
Similarly, increasing N does not help improve the Nakamoto Cx at all. The coefficient is purely a function which reflects stake decentralization in face of the consensus rules, so see the paragraph above.
In other words, it makes no sense to try to improve decentralization via an expanded set. If those aspects are truly a concern, they should be first addressed within the current validator set.
As for the 1st point, Geographical Distribution of Nodes, an increased N would yield even more low income validators joining the set, since stake decentralization isn’t being efficiently or directly addressed. These low income validators for the most part run nodes in two setups: at one of the large cheaper providers of lower quality (e.g. Hetzner/OVH/Contabo), or at home. While the former setup actually decreases geographical decentralization, the latter would indeed improve it. But the question then becomes, does the network benefit from nodes running in consumer HW in people’s basements, on likely a single uplink and no power redundancy?