I'll bite.
Regarding your assertion:
I'm not sure that follows. TCP/IP's primary design point was avoiding centralized network failure in a nuclear war, so you can't strongly claim that it is responsible for heirarchies (i.e. Star) we see in modern topology -- that's likely due to other forces like cost optimization (eg branching structure is a common feature of such optimizations).
Now this isn't to say that TCP/IP is efficient for all decentralized networks. And I think there are quite a few points that make mesh networks challenging... But you could also argue that for the interplanetary network protocols as well, so I think it's not so much the concept of decentralization, it's more the specific constraints of mesh or interplanetary that are unsuitable for TCP/IP.
Regarding the quote:
This argument seems to be partly based on economies of scale and how they are currently applied. But internet economies can be platforms as well as products.
Platforms have a strong tendency towards monopolies (i.e., Apple App Store, Uber, VISA). But even in platforms, they may centralize wealth by treating customers as resources, or they may distribute wealth by treating them as partners.
As we race towards the so-called "gig" economy, I think a key differentiator in platform providers will be the extent to which they grant power to their users as partners in creating value.
This changes the classical idea of monopoly quite a bit as the old concept merely controlled a market for maximum exploitation of its consumers. If consumers are given back some power in terms of profit sharing, content creation, value creation, the Internet could be an extremely vitalizing force.
You see this in spates: Etsy, Steam, Heroku.
Maybe greater collaboration will be necessary to drive increasing efficiencies. The key point might be "central management" vs "empowerment" rather than platform, product, monopoly etc.