Jesus, I don't know how I could explain it any better to you, but I'll try once again 
...
It is the lazy devs answer if the synchronized simulation is creating a huge problem in itself... lag.
The synchronized simulation is causing "lag" as in "stutter" or "halt" if a datapacket between on of the players is delayed for too long. The same thing would happen in a server-client setup.
Nope, you are the one missing the point. I'm not saying whether the game would or wouldn't stutter - I'm saying there is no advantage to using p2p over client/server from a latency point of view. This is what we were told when the game was released.
But it is decreasing the latency. As you posted yourself already, with the peer-to-peer model, every participant has a direct connection to each player and thus the best possible latency. As opposed to a server client model, where the effective latency is the sum of the latency between one player and the server and another player and the server.
Yeah, right up until you stick 350 ms artifical ping onto it!
Again, the netlag is just an arbitrary value that enables players to play with each other fluently even if they have a high ping to each other. The netlag value is completely independent from the used network model. The peer-to-peer model just decreases the latency between players (when compared to a server-client model) and thus increases the range of players who can play with each other fluently.
If you have 3 players, everyone with 200 ms latency to each other, their effective latency in a server-client environment would be 400 ms, while in the peer-to-peer environment it's just 200 ms. It's not that simple of course, since the games with a server-client environment often use a hybrid technique and thus reduce the effective latency. But in the end the direct connection is always the best possible latency, naturally.
