Yes, you're right, the distribution need not be uniform. The assumption is that the distribution is such that the probability that all the constants are within the Goldilocks range is very low (not even necessarily that the constants are fixed). But yes, either way, my answer is that there's no reason to believe this assumption.