First, one should not apply statistics retroactively until you hit large numbers. Second, I am talking about design goals, not track records. You don't evaluat

First, one should not apply statistics retroactively until you hit large numbers. Second, I am talking about design goals, not track records. You don't evaluate the safety of a large system by running it a lot of times and see what happens - you design the thing towards a goal of safety. So this involves designing each component to the specific failure rate, and testing individually, then then testing together, documenting every step, to provide traceability of every change on every component. The comparison to historical data on Falcon is not entirely valid, because we don't know the design goals of Falcon, etc. Just because Falcon 9 is human rated, doesn't mean Falcon Heavy is human rated, etc.