"Collective rationality" by itself feels a bit circular because it doesn't come with an absolute way to measure how rational a conclusion is that doesn't ultimately loop back to
what we've concluded now.
Perhaps if you used the idea to compare against some abstract scale (like, for the sake of argument, utilitarianism) it would be an interesting conclusion but without it it feels like "what's popular is right because what's right is popular", and popularity really isn't a great measure of quality[1].
A big part of the problem is that this vaguely teleological process, to the extent it exists, is almost certainly not monotonic. Even if we agree to, hmm, eventual consistency, it's hard to believe things are always improving and never getting worse. But this means that the idea all of a sudden has surprisingly weak predictive power. It's easy to analyze history and come up with an argument for many trends we like, even contradictory ones. We can say that, hey, there's this general trend, but we took a step back in the 1950s and we took a step back during the War in Iraq and hey, maybe we're taking a step back now, but the global process exists.
Honestly, it reeks a bit of rationalization.
A concrete example I've thought about: freedom of speech. It's a value I, like many others here, hold dear. And you know, I can make a pretty strong case that we're moving towards more and more freedom of speech both legally and culturally. The government censors less and less, our neighbors tolerate more deviance and obscenity and lots of taboo topics aren't taboo any more. Of course, some new topics become taboo every now and then, and people constantly clamor against freedom of speech, but on the whole we're getting freer and freer. On the face of it, this is a plausible argument.
But I've heard exactly the same reasoning but for ideas that, naturally, are at odds with freedom of speech, like certain notions of "human dignity" (which makes certain things unsayable). And hey, they can build a case just as well!
And perhaps I'm just seeing signal in noise. Taboos change but instead of progressing it's a cycle: old words that have been taboo for too long lose their sting while new ones acquire it. We can actually observe this with words describing less-favored positions like "secretary". It used to be that "secretary" was a perfectly reasonable job description; now we call them "executive assistants" and while "secretary" isn't a taboo per se it's certainly jarring. (There's a name for this cycle, actually, but I forget what it is.)
Honestly, on the whole, I think any strong view of "collective rationality" is, well, too strong. (Just like the efficient market hypothesis—a good point of comparison, in fact.) More importantly, regardless of how true it is as a general trend, I think there is way too much noise to use it to come to specific conclusions. Even with a general upward trend, the variance in what is and isn't taboo is going to dominate unless we have some other scale to value it on.
And, gratuitously, I'm going to link to my favorite of PG's essays which is pretty relevant to the topic if somewhat orthogonal to my point: "What You Can't Say"[2]. It's worth a read.
[1]: Second time today I'm linking to this comment, because it covers an idea I find quite powerful: https://news.ycombinator.com/item?id=10567962
[2]: http://www.paulgraham.com/say.html