Having to list the tokens to revoke is a problem in several ways. Obviously publishing the tokens themselves allows attackers to use them before relying partie

Having to list the tokens to revoke is a problem in several ways. Obviously publishing the tokens themselves allows attackers to use them before relying parties get the revocation notice, so publish a token hash instead. But even publishing a token hash is annoying because you then have to record all the claims + token hashes issued, and that's a failure point. Instead you'd want to revoke _subjects_ (`sub`), but since tokens don't have to have those... (In enterprise systems though your tokens should name subjects.)