how do you handle rate limiting for busy apis

Hey everyone, I want to talk about rate limiting for busy APIs. We've been using a basic token bucket setup, but it's struggling when we get swamped. What have you all found that works well for handling lots of requests at once? We need to keep things running fast and make sure it's fair for everyone. Any ideas on things like distributed rate limiting, or maybe sliding window counters? What tools are you using?