I feel a long post coming on so I will split into two and answer SAS and finalchecksplease first - cos that's easier!
SAS, I agree that the forecasting of icing is not particularly accurate but that does not really matter. If we are going to depart and fly at sub-zero temperatures, if by day I can avoid scattered clouds. If by night or where the cloud is unavoidable I will ensure that I have my "warm" band 1000' amsl and below. That is very predictable as its related to sea temperature, which only changes very very slowly. So if light icing is forecast but it becomes more than my aircraft's limited icing clearance can accept, I just have to descend into the warm band. Its a nuisance and reduces com range etc but not really a big deal. It would be a brave person who departed without an escape route into sub-zero cloud reliant on the fact that only light icing was forecast (brave = daft!).
So I maintain that rotor deicing does not improve safety, and unless its used cautiously (with an escape route) one day it will all end in tears. The fact the the RFM does not mention the need for an escape route is negligent IMHO.
finalchecksplease - you have not been keeping up with your reading of pprune! JAR-OPS3 used to allow Performance Class 2 (PC2) over a hostile environment (N Sea) only with conditions and only until end 2009, after which PC1 would be required. PC1 offshore is not realistic so JAR-OPS3 has been modified to allow PC2E which is Performance Class 2 Enhanced. PC2 had exposure time between the takeoff /landing decision point and achieving Vtoss/landing (exposure meaning you might crash or ditch if the engine failed during this time). PC2E uses manufacturers drop down data, deck edge clearance data etc to allow mass adjustment so that the exposure time is nominally zero. So its not PC1, but its probable (not certain) that with an engine failure at any time there will be a sucessful outcome.
HC