That seems too simple to me. Can they afford to test enough rockets to empirically validate an acceptable failure rate for high-reliability components? If it should fail 1/500 flights, for example, how could they validate that in testing?
ULA has over 100 launches without failure; SpaceX's record is around 19 out of 20. My understanding is that SpaceX will need to reach ULA's level of reliability to succeed, especially to launch humans.
ULA is triple the cost of SpaceX and they have an $800 million rider with the DOD for "at moments notice" launch. That makes them almost quadruple the price SpaceX. [1]
[1]https://arstechnica.com/science/2016/03/ula-executive-admits...
My point is that there are other methods of assurance; one of them is additional investment in reliability, something SpaceX is trying to avoid.
> ULA is triple the cost of SpaceX ...
Perhaps that's how much sufficient reliability costs. In much of engineering, and at least in IT, high-reliability is one of the most expensive requirements. As a simple example, compare the cost of an unreliable consumer ISP service ($30/month for maybe 15 Mbps?) and a reliable T1 circuit ($300/month for maybe 1.4 Mbps?) - you pay 10x for 1/10th the speed.
From my seat in the bleachers, it will be interesting to see if SpaceX can deliver high reliability for significantly less than ULA. Good for them and for everyone if they pull it off.