Engineering is still less of an exact science than most people believe it to be...
:/
Engineering is exact science. The question you need to answer is: How much would you be willing to pay for things engineered to perfection?
Part of engineering is making assumptions. Those assumptions are very useful in keeping costs down, but of course have the potential to increase the margin of error when calculating safety factors.
Of course, when you do have to make assumptions, or replicate boundary conditions that are not exact (again due to cost/complexity), you err on the side that will be worse. As an example; you might be testing a component that is installed in an aircraft. However it is not cost effective to get an aircraft frame simply to install your component and test it. So you could consider what the modes of failure would be for this component. Is it purely structural? If so, if you rigidly attach it then the loads going through your component will be higher than if it was installed in the aircraft, as your rigid attachments will not "flex", thereby they won't absorb any loads. In real life, the aircraft would "flex" (even if by a tiny bit) and would absorb some of the energy.
What if you have 40 different components that are similar? Do you test all of them, or try and determine the worst one in terms of its likelihood of structural failure and test that one?
I know in a lot of the safety critical industries they do exactly that. Make assumptions, err on the side of caution, determine worst case, etc. All to save cost of course, but without adversely affecting safety.
Another example of how cost drives industries is that of automotive vs. aerospace. In the automotive world, you will crash the entire vehicle no problem. Do you think they crash entire aircraft just to test their performance in a crash? Even if Boeing/Airbus decided to do that, what about the millions of options in each aircraft? Each airline chooses their own seats, lavatories, business class, first class, sometimes engines, etc. Same stands true for automotive as well, do you expect them to test every permutation of every option? Of course not. They will usually test the heaviest option (in terms of crash testing that will tend to be worst case), or whichever option(s) yield the worst results in terms of the considered failure modes/test criteria.
All of the above introduce a statistical uncertainty because you have not physically tested everything, or because your boundary conditions will never exactly match real life scenarios , or because the scenario might be slightly different, etc, but it affords all passengers a
reasonable level of safety and ensures the industry lives to both make money and serve customers.