I play video games, and I can't imagine what the latency would be like over a mesh network. So many hops, so many pieces of questionable cheap home hardware handling my data...
You will end up with a random geographic distribution of radios with different capabilities. The number of network hops will be reduced because, as with skip lists, you find a path to high capacity station, then skip longs distances, then find the final recipient.
In practice this would be difficult to deploy (Can you legally give high power stations? What about RF safety of higher power stations?), but I think the idea is neat.
1. Neither can anyone, because we don't have examples of large-scale usage of mesh networks, and the amount of research that has gone into mesh networks is nowhere near as extensive as exists with ISPs. There's no reason to assume that mesh networks will always have higher latency, especially in places where i.e. Comcast is choking connections.
2. The small-scale mesh networks which already exist are demonstrably adequate for many, many use-cases. Mesh networks which handle text and image web browsing, email, instant messaging, are still extremely useful even if they have trouble handling streaming video and video games.