Opaque fragile systems/institutions dominate.
“Out of the crooked timber of humanity no straight thing was ever made.”
All systems compete against each other for users. I believe opaque fragile systems dominate transparent robust systems. First I believe individuals choose shrouded systems more tightly bounding their rationality. Second I believe even when individuals know a system is fragile they believe they will not be victim to its fragility; the greater fools will be.
First sophisticated consumers like price discrimination while myopic consumers are ignorant they are being discriminated against or unwilling to commit the time needed to exploit the system. Information asymmetry is a feature of the system not a bug.
See “Shrouded Attributes, Consumer Myopia, and Information Suppression in Competitive Markets” http://aida.wss.yale.edu/~shiller/behmacro/2003-11/gabaix-laibson.pdf for more detail.
Second sophisticated individuals, even when they perceive the system to be fragile, often subscribe to the the greater fool theory. They feel they will win out over the greater fools, but they do not understand how tightly bound their rationality has become due to the layers upon layers of shrouding. The myopics of course are largely ignorant of the risks.
This is a very pessimistic view that offers no solution to the problem of system fragility. I think though most solutions to fragility only create larger equally fragile systems.
I have a few problems with this theory.
First is that the boundary between "sophisticated" and "myopic" consumers is not only artificial, it is either patently false or non-falsifiable. You could only verify a consumer as sophisticated/myopic based on whether they favored price discrimination and transparency. But it is almost trivially true that nearly all consumers will sometimes favor these institutions and sometimes not, so "sophisticated" and "myopic" don't neatly divide agents. You could, alternatively, say that it is activity and not the individual agents that the kinds apply to, but then you could divide any set of agents that way, regardless of how they were composed or what activities they took, which would not be enough to use the kinds to establish trends. This same criticism can be leveled against "fragile/non-fragile" systems/institutions.
Second, it's not clear how this sort of division is useful. As you point out, it offers no solution to the problem it poses. While it can be true that some problems may be beyond solution by current means, our means includes our current knowledge of the problem. A well-developed theory of sophisticated/myopic consumers and how they relate to institutions would give you some idea of what would be needed to make a system less fragile or consumers more sophisticated, even if those means were currently impractical. For instance, if system size (in terms of people, dollars, influence, or some other metric) were the issue, it would be fair to posit that making smaller systems would make them less fragile. (I picked size only because you mention it specifically.) But this theory seems to directly suppose that this is just the way things are.