Predicting what experiments have potential for progress

How to know what experiments have the potential to find absolute limits to the modifiability of theories? Let's start with analyzing cases where experiments did find absolute limits to the modifiability of theories, and contrast them to cases where apparent anomalies were not absolute falsifications. One of the paradigmist's favourite examples is when only a third as many neutrinos from the Sun as predicted was observed, some suggested that the first law of thermodynamics was wrong while others suggested that the Sun was cooler inside than thought, it was later found that the Sun only produced electron neutrinos but many of them transformed to muon or tau neutrinos on their way to Earth (implying that they had mass). The paradigmists who try to generalize that to all shifts of theories deny that the Michelson-Morley experiment falsified the aether, and try explaining it away by claiming that aether theory was never useful but relativity is very useful. That ignores the fact that aether theory did have some use (it was the only theory known back then to explain the wave properties of light, with some practical applications). While those applications pale compared to those of relativity, so would the applications of relativity pale if a theory of everything appeared. The key difference between the neutrino transformation example and the aether falsification is that in the case of neutrinos, the missing neutrinos could be calculated at a predictable mathematical percentage of neutrino transformation. The Michelson-Morley experiment, by contrast, discovered that the speed of light is the same in all directions, which due to the movements of especially the Earth could not be explained by a increase/decrease of the speed of light at a fixed ratio. So in order to find the absolute limits to the modifiability of theories and get science going again, experimentation must be focused at things that cannot be explained by generalizable ratios but instead either should be expected to or do vary in many ways at different measurements.