Suppose the client follows a link to old-slug after the slug has been changed to new-slug. The server responds 301 → new-slug. The client caches that redirect, so that if you request old-slug it will immediately take you to new-slug without querying the server.

Then the object’s slug is changed back to old-slug (perhaps the change was made in error). Now a request to new-slug produces a 301 → old-slug. This likewise is cached, and a client may new be stuck in an infinite redirect loop.

I’m not sure if this is actually what browsers do; they might detect the loop and decide to throw away their cached redirects. I haven’t tested it; but I wouldn’t count on it.