Robust GraphQL Rate Limiting Middleware for Flask

Design a comprehensive rate limiting middleware for a GraphQL API using Flask that implements dynamic token bucket algorithm with per-client tracking. The solution should handle concurrent requests, provide granular control over request rates (e.g., 100 requests/minute per API key), and include automatic client throttling with exponential backoff. Include proper logging, Redis-based tracking, and graceful error handling for overloaded clients. Demonstrate how to integrate this middleware with existing Flask-GraphQL applications.