I think that a lot of it is based on the general premise that market actors are nearly perfectly rational (which is the general premise in a lot of libertarian economic analysis, and, more broadly speaking, the Austrian school), which just doesn't hold up to practical observations.
> What Austrians and their fellow travelers can argue is that it’s not the rationality of market participants that matters, but the institutional context within which they act. In other words, rationality is not a feature of the individual choosers but of the market as a whole. Even if people make “mistakes” by not acting as the strict model would suggest, they will receive feedback from the competitive marketplace that will demonstrate their errors and give them the incentive and knowledge to correct them. Those who can recognize their biases and correct for them will do better than those who can’t, and markets enable us to do that when they are genuinely free and competitive. This is what Nobel laureate Vernon Smith calls “ecological rationality.” Even if individuals are irrational, the system as a whole produces rational outcomes.
IMO that is also not consistent with observations of real-world markets. Part of the reason, perhaps, is that such feedback would only exist if a relatively small part of the market is irrational, while the rest of it is - then the irrational part is, in a sense, "punished" by the rational part for deviation. But if the entirety of it is irrational, then it's not at all a given that acting irrationally would actually produce a worse result for that particular actor - it may well produce a better result due to irrational behavior of other actors.
It could be argued that so long as someone in the market is worse off, they would "correct" their behavior, and that would cause a cascade of punishment/correction for other actors in the chain. The problem with that is that it's not an instant process, and the time scales on which it works - especially when the actors that get the short end of the stick also have relatively little market power individually (and so often would need to organize to effect any meaningful pushback), and so this process is simply too slow - it doesn't actually produce an observable result on a scale on which individual humans feel it, and so it's insufficient to affect their behavior. And by the time it does happen, other irrationalities may well be introduced that may often mask the effect, or simply make it irrelevant.
(FWIW, I'm generally very suspicious of the Austrian school precisely because it actively rejects empiricism in economics, and substitutes hacks like "praxeology" in its place.)
