Local affiliates can cross license content directly between each other, or from NPR, PRI, BBC, American Public Radio, and Public Radio Exchange, etc.
Local affiliates are not simple repeater stations, as you seem to believe.
When I visited Oregon, I noticed most of the shows were nationally syndicated, and I suspect that might have to do with how much money they can get from their local regions.
Unfortunately, just saying "We produce Car Talk/Fresh Air/whatever" is going to be enough to salvage the affiliate as presently composed; if they have a show of value, they'll have to make it available independently and will just be the Fresh Air company, not WHYY in Philadelphia.
A 100W trasnmitter can cover a small town, 1 kW - 50 kW is (very roughly) the range of many larger FM stations, and can reach an area 100-200 miles in diameter, though usually less given geography and competing stations. Doc Searls (an old radio head himself) fairly frequently posts stats about various stations, and their reception areas.
The best thing about broadcast is that more listeners don't cost you, and there's no last-mile problem. It's particularly well-suited to mobile uses (especialy cars).
The downside: what's on is what's on. This increasingly is my frustration, and the more I'm used to on-demand services, the less viable I find this. There are alternatives, such as a DVR type service which records over-the-air broadcasts for later playback, and there are some concepts along these lines, though most prove pretty fickle and annoying, with little uptake.
I suspect there'll be a place for radio, though that may shift a bit with time.
The stream is already free-as-in-beer, what is somebody going to do to it?
I want to be a good seed, but not if it costs me hundreds of dollars a month.
It's possible to make a radio out of a couple of crystal diodes. A tuner and FM receiver are slightly more complicated, but a 9-volt battery or very small solar cell will run a receiver, and transmitters, as noted, are pretty bog-simple.
Another quite probably underappreciated element is security. It's exceptionally difficult to hack into a radio, or into a radio transmitter, through the radio itself (there are of course other pathways to hacking a modern computerised radio broadcast operation, but there's no reason such an operation need be computerised).
Bittorrent is wonderful stuff, it's a brilliant hack, I use it myself. It's more useful at distributing large files than small ones, and benefits greatly by large audience (saturating many sources). As part of a samizdat information distribution system, it's great.
But you need servers, recipients need computers, you need a network, with last-mile connectivity, everything needs power, there's security issues, there's authenticating who's what, and more.
Sure, you need all that stuff, but you already have all that stuff.
Rather than argue the point further, as you seem bent on denying the point, I'll suggest you list out as many counterexamples of your claims as you can.
I'm up to 20. Care to beat that?
Radio is a technology that connects communities much more than television or Internet access can, especially during times of natural disaster or crisis.