That's a good point. However, it suggests to me that one should publish the code for an additional reason:

1) if we acknowledge that paper production is the wrong way to do science, perhaps papers with attached published code are a step in the right direction. A critical mass of papers that are easy to reproduce because anyone can execute the code attached with them could, hypothetically, come to dominate the zeitgeist and push out papers with grandiose but unverified claims.