That said, I worry about the future we are creating by entrusting so much security and traffic to a single point of failure.[1]
It seems to me that CloudFlare is positioning themselves as another Google or Facebook, where a key feature of their business is that they get to track the web history of a large portion of internet users. Much like Google gets to have a lot of my email when other party is @gmail.com, CloudFlare gets click histories by people using their CDN and caching/filtering servers.
While I don't really know anything about the motives and personalities behind the company, I can give them the benefit of the doubt for now. Unfortunately, their motives don't matter - in a world where "national security letters" Prism, XKeyscore, and the like exist, CloudFlare's motives may not matter.
Even more concerning is that while it my hard to avoid Google's tracking, it is at least theoretically possible. With CloudFlare (or any similar service) we are stuck with a situation similar to tinyurl/t.co/bit.ly [2] where content is hidden behind serves from which you have to request the real URL (or in this case, the content itself).
Don't get me wrong - SSL becoming significantly more common is good regardless of what else is going on, and CloudFlare still deserves a lot of praise for advancing a problem that has been so resistant to progress in the past. I would even agree that CloudFlare's caching services and security protections are (very) good engineering techniques that we should be using. It just seems like everybody is (yet again) setting up a single point of failure that will suddenly have very significant consequences the minute somebody with real power decides they want those very-revealing server logs.
[1] I'd be the first to admit I don't have the best understanding of how CloudFlare works; corrections to any misunderstandings I may have about their business or tech would be greatly appreciated.