Suppose someone implemented a list which for any IP address returned
127.0.x.y, where x.y mapped into how spammy the list maintainer
thought the IP address was. It could be used as both a whitelist ("if
the score is low, accept") and blocklist ("if the score is high,
reject") as well as for other purposes ("if the score is low but not
low enough for auto-accept, subtract 2 from the SpamAssassin score; if
it's high but not too high, add 2; if the list says the IP emits much
but not all spam, run these extra checks").