Protein origami master?

This story puzzles me. Allegedly the protein Hsp90 keeps other proteins from mis-folding even when there are minor defects in the protein due to minor mutations. Too little Hsp90 and the little changes tend to be expressed (almost always as defective systems) rather than suppressed.

That's a nice method for achieving largish jumps in characteristics. Stress a population of organisms, reducing the Hsp90 or demanding so much protein creation that you outrun your supply (I wonder about radiation environments?) You get lots more previously hidden mutations expressed, almost all of which are damaging (but we take it on faith that some can be pathways to beneficial changes). The only real example of a successful change is fish whose eyes change size (smaller or larger) under Hsp90-blocking, which might emulate the stress in the slightly too non-conductive water of the dark cave some of them live in. If you are already halfway there you should show a more rapid tendency to lose the genes for eyes.

Except--how does this wonder protein know how each protein is supposed to fold? If there were a general folding pattern common to all proteins, then the normally suppressed deviants would always be abnormal and not the wave of the future. For if some deviant folding were a new pattern for future proteins, then from then on there would not be a general pattern common to all proteins. If there are no general folding patterns, then how in the world would Hsp90 do its hypothesized job?

Interesting work, but I think something is wrong with the model of what Hsp90 does.