Just because it's always worked in the past doesn't mean it will always work in the future. I've actually spent some time thinking about this in the past - a big hidden assumption of science is that everything is knowable, which we know is not strictly true, and I'm not entirely confident that it's even approximately true.