The routing algorithms and therefore the required network topology are a matter of research. You can’t really do the correct calculations without knowing what the eventual routing algorithm will be like.
The problem with Fyookballs’ argument is precisely that. He made erroneous assumptions about the routing algorithm and got erroneous results.
Another problem with doing the correct calculations is that a network like LN is a living social network. Not a purely mathematical construct that’s easy to analyze. The actual network provides incentives for participants to intelligently form new links that improve the routing performance of the network. Fyookballs’ assumption of a random graph is hence invalid.
Any argument that’s based on purely random graphs is therefore flawed. However, I don’t have a clear enough idea of what the actual network will look like to be able to model it for a proper analysis. I only know enough to point out why Fyookballs’ analysis falls short.
In other words, while I know enough to show that Fyookballs’ analysis is flawed, I lack the knowledge to be able to perform a proper analysis.