the documentation actually says that script execution limits are tied to concurrent runtime slots, not just api request counts. when you see 429s despite available headers, it is usually because the platform has exhausted the dynamic worker pool for the specific tenant region. the current configuration in Europe/Paris shows 120 active slots but only 80 are utilized, yet the load test triggers rate limiting immediately. is there a hidden throttle on the scripting engine for multi-tenant apps or a misconfiguration in the queue capacity settings?