Hmm, I'm tempted to think this is an analogue of Muphry's law, where someone correcting someone's grammar is likely to make grammar mistakes himself. (Or, at least, that we are much more likely to notice if he has.)
Getting a large enough number of samples is not enough to show that a correlation implies a causation. It merely demonstrates that a perceived correlation is real and is not the result of random sampling error. Take a few samples, and you'll see a strong correlation between wearing a cast and having broken bones; take an enormous number of samples, and you'll prove the correlation is real, and you'll show the precise magnitude of the correlation with small error bars; but you'll never prove that wearing a cast causes broken bones. To prove causation, you need a causative theory, and evidence that distinguishes it from competing causative theories.
Though perhaps you are thinking of situations where the only plausible causative theories are "A and B are unrelated" and "A causes B".
Getting a large enough number of samples is not enough to show that a correlation implies a causation. It merely demonstrates that a perceived correlation is real and is not the result of random sampling error. Take a few samples, and you'll see a strong correlation between wearing a cast and having broken bones; take an enormous number of samples, and you'll prove the correlation is real, and you'll show the precise magnitude of the correlation with small error bars; but you'll never prove that wearing a cast causes broken bones. To prove causation, you need a causative theory, and evidence that distinguishes it from competing causative theories.
Though perhaps you are thinking of situations where the only plausible causative theories are "A and B are unrelated" and "A causes B".