[leoedin]

To what extent is that true when the cause of the fatalities is the technical design of completely unrelated systems?

If one space agency built a rocket which always immediately exploded after launch, and another space agency built one which always worked, you could say the odds of failure for the next astronaut was 50%. But of course the two rockets are essentially unrelated. The chance of success of each rocket is a function of design, engineering process, organisational culture of that organisation.

Telling the astronaut strapped to the top of the explody rocket that there's a 50% chance of exploding is actually less help than no estimate. Because actually there's a 100% chance of them exploding. An estimate is only as valuable as the assumptions that drive it.