IME, at industrial conferences, your pitches get accepted if you are friends with the organizers, or if someone who is friends with the organizers vouches for you to the organizers or co-presents with you. That's about it.
2) It is wildly not the case that all speaking gigs begin with having an in with the organizers, just like it is wildly not the case that all jobs or contracts are awarded to nephews. Pitching is an art. Doing it well will raise your hit rate. Failing to do it at all, while either waiting for the world to recognize your genius or bemoaning the unfairness of it all, is an incredibly effective method to not get speaking gigs.
Our keynotes were entirely invite only (O'Reilly often has sponsored keynotes, we did not) and Kate was invited based on her position in the field and because she's familiar with speaking at O'Reilly events, so we knew she would be good. All keynoters were invited on these merits (so yes, being known to O'Reilly or the chairs helps, but overall stature is just as important).
For regular sessions, a handful of people we knew would be good were invited directly into the program (perhaps 10% of the final program), a much larger number were invited to submit a proposal for review (perhaps 50% ended up being these people), and then about 40% were entirely from non-invited proposals, including several speakers who were new to speaking.
The most important thing to make it through our program committee was to get the proposal right. A good title, a good abstract and, in most (but not all) cases, having a track record in the discipline (even if you're not 'famous' for it). We did take a risk on several 'unknowns' to keep it interesting and in most cases, it played out well.
This is just how we did it, but it's not necessarily the 'best' way to do it (I don't think there is a 'best' way). Some conferences, for example, anonymize proposals, don't use proposals at all, are 100% invite only, or follow unconference formats. It's great to see a real variety and not everyone just following one approach.