Simply knowing that it increases the chances that the startup will succeed should be sufficient to understand where the fixation for multiple founders stems from. True, it's not required 100% of the time, but, a-priori, how can anyone know whether this is one of those times. There are a hundred well defined things that can contribute to the failure of a startup (in addition to the thousand less well defined things) - one thing that YC brings to the table is identifying that list of 100 things that can kill a startup so they can be avoided. Coming in with at least one cofounder brings the list to 99. Incorporating in Delaware to 98. Having someone who can code 97. etc, etc, etc...
Head to http://pmarchive.com
For example, if you have two founders, then one can go out and do fundraising for a month while the other keeps the company running. If you had only one founder, you'd be dead in the water for that month. (Even if you have employees -- things are going to decay if all of the founders are distracted.)
Also, I've never known YC to play founders against each other.