When the Saab 340 came out first it didn't need any icing to get into mischief. That got sorted pretty quickly but it was entertaining.
Re Roselawn, ATR 72 etc, The event gave rise to substantial changes to §25, and added to the Appendices to the part; runback icing became a big deal, as did SCLD. (Both off those issues happen to be well modelled by FENSAP-ICE if anyone from an FAA ACO is listening... ). The rules frown on roll reversal prior to the stall however defined by the manufacturer, and for a small period below the stall reference speed. Getting reversal is usually a pretty lousy bit of design in the first place. A prop powered plane may get a pretty neat wing drop at the stall, but the ailerons should not get a reversal before the margin after the stall.
A B737 managed to get a recorded aileron reversal, happened to be about VNE+120KCAS at the time and was fixed by the wings departing the fuselage with unfortunate results. doing an inflight realignment of all AHRS platforms in a thunderstorm is part of nature selection process at work.
Wasn’t the ATR a reversal situation or a case of aileron snatch. I thought it was the latter. Will have to re-read the Indonesian 737 report as I don’t remember that issue (or is it hidden away in an annex).
Would like more detail on the Saab issue.