When our Elasticsearch cluster was recently overloaded, our API avoided more than 23,000 retries in 15 minutes, load that would otherwise have hit the struggling cluster. Here's how a simple token pool lets us retry when it helps and fail fast when it doesn't.