This is an interesting idea. If this gets widespread-enough adoption, and if non-metro folks can, on average, find a nearby peer for $whatever they're looking for (and those are very big "if"s), I still imagine some problems might crop up:
Determining "distance" is quite tricky, and the trickiness goes up substantially the further onto rural/less normal latency distribution networks/ISPs you get--and that's just in the first world. If latency is the main signal used to inform distance, you could end up inflicting a pretty punishing upload-bandwidth cost on users/ISPs because of, for example, a fortuitous peering agreement upstream.
There may also be a significant penalty paid for "medium distance" non-hub-routed peers: situations in which the presence of many uploaders congest the last-mile/backhaul infrastructure of small-scale providers that are least capable of managing the unexpected surge in consumption (least capable as in: most congestion-prone infrastructure and fewest technical solutions on hand to mitigate unexpected long-term congestion).
And in places where small-scale providers don't control much of the market and mobile networks are the main bandwidth providers, expect a swift no-warning traffic block. There are ways to circumvent those, but they get expensive (in engineering time, user experience, and user bandwidth) fast.