There is a difference between "not always reproducible" and "not reproducible at all" and even when you can't reproduce something, you usually have at least an idea why. What's the success rate of reproducing cold fusion, 0.0001 %?
You cannot prove a theory per se, but you can usually tell where do you expect it to be wrong. Valid theories have inner consistency, meaning that they don't deny themselves. During the history of physics there was always something a current theory couldn't explain, so another theory had to be devised. The difference between cold fusion and say black body radiation is that almost nobody was able to get cold fusion running and even when somebody did, the results were far from being convincing. Black body radiation and subsequent Planck's discovery of quantizing of energy was much bigger breakthrough which completely changed our understanding of the Universe, yet nobody protested against that, because he Planck presented a conclusive proof. Do cold fusion pioneers have a conclusive proof that it works? If their experimental data has been tampered with, why don't they just try again at another research facility or why don't they publish the way their experiment was conducted so somebody else can give it a go? I'm sure there will be someone willing to fund such a research if their theories and finding have merit...