The buttons suck - we knew that a lot of users hated them (it adds 4 * 20 additional HTTP requests to the main page), but they were worth too much traffic to do away with. Social network referrals likely surpass search engine referrals for a lot of blogs and web sites. They have steadily become an important web traffic navigator.
If somebody can figure out how to make those share buttons prettier and more efficient there is probably a product in that. I would guess that most blogs would love to drop the grid of share buttons that can be found on every post.
Otherwise I totally agree with not going for subdomains. We setup each property on a separate domain and initially had some on subdomains. The subdomains didn't rank at all and didn't help our PR or SEO. As soon as we switched each property to a separate domain our search referrals rocketed. For eg. you can now find a crunchbase link within the first 5 results for the name of a startup, while similar records for posts that lived on subdomains wouldn't rank at all. We had around a dozen different domains and frequently linked between them (for eg. each post would have multiple crunchbase links), and it worked really well for search ranking (search engines are ~40% of crunchbase traffic, IIRC) It shouldn't be like that, but it is.
I found that quite nifty and unobtrusive. A smart solution to the problem.
There is a product here - package the buttons together so they only load once, hide them in some way by default, and provide a bunch of analytics. Lots of blogs and websites would pay for that - the current solutions for social button analytics are really bad, or non-existant (you need to know things like click-through rates, which share options are most popular, autohiding some for users who never use them, detect if the user is logged in for each service, how many followers/likes/retweets etc. each user gets from a share so you can identify 'power sharers' etc.) - I don't think that exists.
You can't just keep adding more and more buttons for each social service, and there are some people who will only use, for eg. delicious or pinboard and there are no buttons for them.
Mobile is definitely a problem but the mobile version of the website makes that solveable for the most popular devices.
One thing I'm learning about SEO, is that you have to know your audience. This means you should know what they are searching for, and optimise for that. Likewise if your audience is a big user of twitter (which TechCrunch's audience would be), then use that to your advantage.
Am I wrong?
Can anyone clarify?
If the price isn't too high, someone should just start selling crapblockers that make everyone happy. I guess the challenge is that you can't price a crablocker until after you've inundated someone with crap for measurement purposes.