My understanding is this (and those of you that know feel free to correct me):
When engineers design something, they have to work out/guess what loads the component is going to be subjected to. The magnitude of the safety factor is dependant on the confidence of the accuracy of the load cases and the level of modelling. So in effect I would expect the SF in the S61 to be higher than that in the S92 (through better computing power if nothing else).
If my point above is generally true (ie that modern designs have smaller safety factors due greater experience) and that FMEA use history to determine probabilities of failure, am I the one to see the flaw in this? If the historical data is based on older components with larger SF, the probabilities cannot be applied to newer components.
I am way out of my depth here and hope that a proper designer would put me right.
But please don't forget it is not always easy sitting in the middle of the north sea on a dark night thinking that we are driving a prototype (ie a newish machine) where the design philosophy is based on potentially shaky probabilities.