The hypothesis is that as we increase the block size to accomodate more transactions, then the orphan rate increases due to higher block propagation time across the network of nodes (and that time t is in the equation for the orphan rate). Thus we say there would be an equilibrium point at which the level of transaction fees the market would bear meets the cost in wasted hashrate. That is Peter R's (and Bitcoin Unlimited's) thesis (in their whitepaper).
That's kind of ridiculous, indeed. Fees drop like a stone when you relieve market pressure, much much earlier than when orphaning rate through network propagation becomes a worry for miners. It can only happen when network propagation times start to be of the order of the bloc times, at which point, the network becomes totally useless, because if you saturate the HIGH SPEED links between miner pools with blocs, no normal user can ever hope to download the bloc chain.
If you get orphaning because propagation delays between miners is of the order of 10 minutes to get ONE BLOC through, you can forget downloading the whole chain on your PC with a lesser link through a P2P network.
Simply ridiculous. By the time his equilibrium is reached, nobody has a copy of the bloc chain any more except for the miners with their 100 GB/s links between them. And they won't have any transactions to put in their blocs.
This is not a PoW system any more, but a Proof of network link.