I've never had a paper rejected from a conference, and I've seen some pretty shitty ones at conferences. I think this statement is pretty field-specific.
That said, in the CS and EE fields at least, the use of 10-12 referees for a short, dense conference paper is a really good idea. It would appear that this works primarily because of... drum roll ...the social norms and incentives in place within these specific fields.
As an example, arXiv is has a lot of great papers but it also has a lot of bad ones. I'm unlikely to go there to find a paper unless someone referred me to a specific one.