Unless it is horrendous, I have erred on the side of caution and just left my collection as is. I am far more concerned with them working well and not having serious defects (scratches, gouges, imprints) than I am some yellowing. To each his own, but that is my route. We will never have real results because even if you, say, took one case today and retrobrited it and another that you didnt, and crushed them and measured how many newton meters of pressure it took crack it, there would be no experimental control since both cases have probably had ~30 years of different storage conditions. You would need to manufacture a test subject, store them in the exact same conditions for a set amount of time, then retrobrite one and do the test. Even then you would need successive tests to get a sample size that would tell you meaningful results.
In essence: no one will be able to tell you with scientific proof that the process makes plastic brittle. You just have to weigh the risks/benefits on your own.