Yes, but the expected value of harm from my backyard pool leaking is, well, I can't swim and I need to resurface the pool liner. Multiply this issue by the prob

Yes, but the expected value of harm from my backyard pool leaking is, well, I can't swim and I need to resurface the pool liner. Multiply this issue by the probablity that it'll happen at all, and you get, well, a filled pool to swim in. But the expected harm from a SNR like this having a leak is decidely not the same. Yes, the leaks may be rare, but the outcomes from such a leak occuring have Chernobyl-like outcomes in their failure trees. The probability calculus still may return that this design is quite fine, given the risk tables, but you MUST run the numbers. Hand waving is not ok here.