It's hard to see how the threads come together but there is definitely something there: faster and faster flow of information allows humans to produce, refactor and combine ideas much more quickly than one mind could.
On the other hand you need some people who don't immerse themselves in the stream too much so they can actually drill down on hard problems.
Another typical strategy is the "girls get in free" nightclub strategy. Although ultimately it might be possible to make money from the market itself, in the early days it's almost certainly the case that whichever side is harder to attract probably needs to get the service for free.
The article doesn't really provide anything specific - oAuth is actually hard to implement and the reason for variations is typically down to both dev cycles and differing opinions on security elements.
We see a lot of openAuth implementations (we provide infrastructure for it: http://www.3scale.net) and we try to help people be uniform, but the spec still has room for interpretation.
I think there'll be convergence - things like http://www.ifttt.com/ will encourage people to the straight and narrow - but it'll take time. Agree with some of the other commenters - not sure how much this article really helps: a detailed laundry list might be better.
Kind of annoying to focus on the "great for the Valley" - surely what's important is it being great for Yahoo. Having a product person in charge again is huge!