"there is no guarantee that your preprint is accurate with the final, published version, so most people will still trust the paywalled version more than your PDF"

I think this is backwards. The definitive version that should be cited is the freely-available one, since that is the version that everyone can read. No one should cite the paywalled version.