Im curious what model/algorithm twitter is using for rate limiting. Doesnt feel like a token bucket.