To voxu:

As we deal with events which are less and less likely to happen, the science becomes more and more difficult. What happens when we get to the stage where we'd be lucky if it happens once in ten years? Then we'd better make sure we're ready to record it in full detail. That one recording, that one data point is gold. There is no realistic hope of getting another one in a reasonable timespan. If you insist on repeatability, then you're saying that science can't use those rare observations, you're making a limit beyond which science cannot operate. This is nothing about the nature of truth or reality, rather it is about the artificial limits that science has made for itself. This is why to go beyond science to see more of reality, you have to adjust the groundrules a little, including being more flexible about repeatability as I suggested.