Open-source is not simply egalitarian. PostgreSQL (for example) doesn't try to apply egalitarian ideals to patch acceptance. Bad code is bad code, no matter who wrote it.
In the ideal, open-source is meritocratic, which also happens to be the ideal criteria applied to selecting conference papers.
> Also, half of the value of tech conferences is networking, not lectures, which is about 'meeting people' rather than 'imparting knowledge' ...
Why do you think these are different things? Some of the best collaborations I've had have resulted from meeting the smart people presenting at conferences, and sharing knowledge.
If conferences break the signal-to-noise ratio by attempting to fill quotas rather than selecting noteworthy papers, then the smart people will stop attending; the social value AND the associated knowledge sharing will be diminished.