NASA wasn’t saying that they know for certain that the Starliner will have a catastrophe on the way back. What they said is that they cannot be certain that the probability of it having a catastrophe is lower than some decision threshold.
Using the russian roulette as an analogy NASA has a revolver with a barel for a million bullets, and they decided they are fine to pull the trigger against the astronauts heads if there are less than 5 bullets in it. But due to nobody really understanding the mechanism of previous anomalies they don’t know how many bullets there are in the barrel. There might be six or more so they are not willing to pull the trigger. (The number of bullets, and the number of chambers is merelly illustrative. I don’t know what is the real number NASA uses.)
Was trying to put this 1/500 number into perspective for myself. It sounds like it is rougly similar to the mortality of having appendicites. [1]
1: https://jamanetwork.com/journals/jamasurgery/article-abstrac...
Therefore the decision to not use it until the causes for all such malfunctions are understood was completely justified.
Problem: "one of the Starliner’s reaction control system (RCS) thrusters did not function".
Why it was not serious: "there are plenty of backup RCS thrusters".
Even if the redundancy ensured a successful return, the causes must be understood, because otherwise at a future return more than one thruster could malfunction and the redundancy may be insufficient.