Well, B31.3 "asks" me to consider a number of cycles that may be, for the most cases in practice, equivalent to one thermal cycle (between the metal temperatures I defined) per day, during a system life of 20 years. I don't think make sense to consider accurate SIFs (and accurate peak stress) calculated considering CA consumed in the first day and having no corrosion in the rest of 6999 cycles. In addition (not related to your question) I don't think it's worth to improve SIF values by FEA and use them in calculation together with conservative values for temperature, defining a range that would appear maybe three times in 20 years.