How about both/and: put paper presentations in rooms that fit a hundred people, but if more than half the committee answers "yes" then move them into a plenary slot! Of course, this would require a RepresentativePC (of all attendees)
rather than an ResearchPC (of active publishing researchers). Or perhaps a different committee could make that call.--JamesNoble.
Are papers for a general audience or not?
I haven't gone to more than one paper session per five conferences I've attended, ever in my life. (But I might be really weird, though I do know a bunch of other people who would agree with me.) What I do is scan the proceedings, look for things that are potentially interesting. I give the paper a quick read. Nine times out of ten, I don't go. Once in a while I go to hear that paper.
Mostly I go to panels, demonstrations, and hang around the edges of conversations with gurus that I'm interested in.
Again, I might be off the charts. What do we know about turnout for paper sessions, as a rule, over time, ... ? -- RonJeffries