The problem with all these debates about "optimal" office layouts is that they're all a series of non-repeatable experiments: someone can always say "Sure, you were successful doing X, but if you'd done Y instead you would have been more successful."

The hypothesis that Y is superior then becomes non-falsifiable: if someone doing Y fails, it wasn't because they did Y it must have been another factor, and if someone not doing Y succeeds, then they would have been better off doing Y. At some point, the argument becomes completely unhinged from any real-world experience.

So really, all I can say is: we have open-plan offices, and we've been successful/productive/kept employees happy, so clearly it is possible for open-plan offices to work. For someone that's convinced that open-plan offices can't possibly be a good idea, and who rejects other people's real-world experiences, what is there left to argue over? The hypothesis becomes non-falsifiable and there's no point talking about it.

To your point about conversations, when people are sick or on leave or working remotely those conversations don't happen and we suffer as a result. We haven't found any replacement for impromptu conversation, or for gaining knowledge through overhead conversations, and so on.

Just as a stupid example (though this sort of thing happens all the time), suppose Chris goes to ask Bob a question, and Alice is setting next to Bob. Bob thinks the answer is that you have to do A, but Bob's wrong, and Alice knows it: the right answer is now to do B. On top of that, Denise, who's also sitting there, hears the answer as well, and just learned something effectively by osmosis.

If Alice wasn't sitting there, able to hear the question, she wouldn't have jumped in, and Chris would have gotten the wrong answer and wasted hours or days doing the wrong thing. Denise also wouldn't be clued in to how things should work either. If Bob had a private office that Chris went to, or it was a one-off IM or phone call, you'd have the same problems. Did everyone get a little distracted by overhearing that conversation? Yes. But ultimately, that productivity hit was worth it, because Chris was saved a ton of time and Denise and Bob gained useful knowledge.

If you can convince people to use something like Campfire where they route all communication through such that people who are remote or momentarily absent are included, I think that can take the place of overhearing those conversations, but it's impossible (in my experience) to convince people to do that when they're working in the same building: people would rather just go chat face-to-face since it's much higher bandwidth than typing.