I guess there can be a million reasons why a "something" might be successful, but it only takes one (sometimes completely unrelated) event to have a very significant adverse effect.
Agree with the human desire to order/model/make understanding of, adverse events, as that is probably the easier thing to do, and it will be very event specific.
Accidents still happen on a regular basis, and sometimes are almost identical in nature. They will also happen on opposite sides of the world, and be entirely unrelated. Doesn't stop them happening though.
There is no doubt the more complex a process, and the more technology that it involves, the likelihood of something going wrong will increase.
You don't see much in the papers about the Monks in the local monastry having a poor accident rate with regard to their ale making activities !
(P'raps we have something to learn from those guys)
I Don't have an answer to the success/failure model - perhaps there isn't one ?
BS