Those are actually really good points... Hiring people is more expensive than saying, "Sure-- come on aboard. We'll see how you do!".
That being said, I'm still not convinced that 10 hours of interviewing is any better than 2 hours of interviewing in terms of predicting success/fit in an organization.
I think 2 hours would be sufficient for 90% of the candidates. That's acceptable when you're in a start-up environment (where you generally have to fight for candidates and the pool is self-selecting to start with), but the 10% at Google could really make a big difference: google's pool of applications is bigger and the "laws of large numbers" come into play.
Also there's something else I haven't thought which is also a matter of "laws of large numbers": you simply can't hire every qualified one (and there would be opportunity costs in hiring the first person to pass the minimal qualifications).
That being said, I'm still not convinced that 10 hours of interviewing is any better than 2 hours of interviewing in terms of predicting success/fit in an organization.