If a log operator publishes a simple rate limit for a log, we can use that information to avoid sending requests to the log that we know will fail. This will improve throughput as we won't be wasting time backing off from failed requests.