when's the right time to implement rate limiting

hey everyone, been thinking about rate limiting lately. we're seeing some increased traffic and occasional abuse patterns. i'm trying to figure out the best trigger points for implementing rate limiting. is it purely based on traffic volume, specific error rates, or something else? what factors did you consider when you decided it was time to put rate limiting in place for your services? any gotchas to watch out for with different strategies like token bucket or leaky bucket?