People also focus their attention.

If I'm at an in-person meeting of 5-8 people, and 3 of them get into a longer discussion about things that are irrelevant to me, and for understanding which I don't have the background, I can zone out and nobody cares, because nobody is looking at me. Everyone is looking at the people talking.

On a video call, my zoned-out face will make the trio feel uneasy; instead of discussing their topic quickly, they'll stumble and try to "talk to the audience", or god forbid, drag everyone else into the discussion. So with forced camera on, I have to look focused while squinting at the other screen, discretely playing 2048 and pinching myself to not yawn visibly, while with camera off, I could get off my chair, stretch a little, or do something actually useful for the company.

I'll be happier with having always-on camera on long videocalls when videocalls start working like real-life meetings: not everyone is looking at everyone else all the time, and everyone can tell who's looking at them in any given moment. The way things are today, I find this constant surveillance to be mentally tormenting, and only activate my camera on the basis of courtesy: 1:1 meetings, some team meetings, meetings with an important higher-up who also has their camera on, etc.

I use people zoning out as a sign that the topic is not a good fit for this meeting, and should be discussed outside of it with a smaller group. That seems to work well to keep meetings quick and on point.

This is fair, but I'm skeptical that these things are actually taken into account in the vast majority of video meetings. Whereas you're right, they're important face-to-face.