And here, exactly, is where the argument begins and ends.
A theory T and some amount of works associates a probability P with using T in some translation of the experimental settings with which P was associated with T.
(using the Universe U as a random variable, which it is not, of course.)
The is the only way to state reliability in any quasi-formal or better manner.
What is EVO other than proving that the Universe is stingy? duh. How many more settings (future/past) can it be reliable in? How can we reliably test EVO this instant?
What if the Universe has a heavy-tail distribution with regards to many of the large scale complex systems within which most of the metrics of Life live?
A theory T and some amount of works associates a probability P with using T in some translation of the experimental settings with which P was associated with T.
(using the Universe U as a random variable, which it is not, of course.)
Those PTU's can get pretty hot.