Incentives, Peter. With the entry cost of setting up a LN hub close to the simple funding of channels, my guess is there will be fierce competition on prices, some of which motivated by purely political, rather than financial, reasons (as in: I, for one, will start a hub as soon as it's practical). With such low-denominator price wars, and low ROI volunteers helping the net nearly for free (as full nodes already do today), I don't see how a high fee level could be sustained.
There was a really interesting talk on "Recharging Lightning" at Scaling Bitcoin Stanford by Aviv Zohar. You can watch it here:
https://youtu.be/3pd6xHjLbhs?t=1h2m21s
He showed that LN fees in fact wouldn't be as low as many people expect because:
- channels eventually become depleted and need to be "recharged." This costs 1 on-chain fee which needs to be amortized over the life of the channel.
- the bigger you make the channel the longer it lasts before needing a "recharge," but then the more money you tie up. There is an opportunity cost associated with this money tied up in a channel (modeled as an interest rate by Zohar).
Zohar shows that the only way fees can be very very small, is if channels are very very richly funded OR if on-chain fees are also very small. Neither of these conditions will hold however (well outside of a highly-centralized hub-and-spoke model where the hubs control all of the LN channels).