No, that's not what I meant. What kills most startups is that they build something lame-- something users see and think "meh" and click on Back. It's not ability to solve problems that makes the difference, but what problems the founders choose to solve.
Friction between founders is somewhat of a threat to startups, but small in comparison with making something users don't care about.
I'd still worry more about an ad hoc group of founders than one that had been friends for a while, but if I had to choose between an ad hoc group with a kickass project and a group of established friends with nothing, I'd probably choose the first.
The thing is, we rarely see ad hoc groups with kickass projects. I think this is because such groups are less likely to build good things.
The great problem with startups isn't a "problem", because it isn't typically the case that startups fail because they have something go wrong. Startups fail because they don't do enough things right.
Following that, the problem joining with a stranger is not that you don't know their quirks, its that it is difficult to find someone with complementary strengths. However, since we hackers are typically drawn to others similar to ourselves, many of those going to YCombinator with someone they have known for a long time will be similarly disadvantaged despite being acquaintances (they will still be two individuals with similar strengths and weaknesses).
Thus what we really need is for two business types to go to YC together, and two hackers to go to YC together, and form two teams, each with one hacker and one business type. Or, better yet, to forcibly enroll the hackers in business courses and vice versa. :)
Friction between founders is somewhat of a threat to startups, but small in comparison with making something users don't care about.