Would it not make more sense for a certain standard deviation away from the mean failure time to still meet the lifespan of the longest living patient? Why a flat 20 years?

Like if your product lasts an average of 40 years with a 2 year standard deviation on failure, if your longest living patient uses it for 34 years then you've effectively guaranteed it will last for life for over 99.7% of users, even if very few will ever last for 54 years.