Why does the /api/v2/analytics/conversations/details/query endpoint starts returning 503 Service Unavailable errors when I push my JMeter 5.6.2 script to 200 concurrent threads simulating WebRTC softphone sessions? I am running this from my Singapore office using a standard enterprise instance of Genesys Cloud. The setup involves a custom Architect flow that routes inbound calls to a WebRTC softphone extension, and I am using the genesys-cloud-webrtc-sdk version 3.1.2 to establish the media connection. The JMeter script uses the HTTP Request sampler to hit the API endpoint after the WebSocket connection is established, with a think time of 5 seconds between requests. Everything works fine up to 150 concurrent threads, but once I cross that threshold, about 30% of the API calls fail with a 503 status code. The error response body is empty, and the Retry-After header is not present, which makes it hard to implement a proper retry logic in my test script. I have checked the X-Genesys-Request-Id in the failed requests, and they all seem to originate from the same API gateway instance. I also monitored the WebSocket connection metrics, and the ice.connectionState remains connected for all threads, so the media path seems stable. The issue is specifically with the REST API calls made to fetch conversation details. I have tried adding the Prefer: respond-async header, but it does not seem to help with the 503 errors. I am also seeing some latency spikes in the ping times for the WebSocket connection, but nothing that would indicate a full network outage. My JMeter configuration includes a HTTP Request Defaults sampler with the base URL set to https://api.mypurecloud.com, and I am using a CSV Data Set Config to feed unique user tokens for each thread. I have verified that the tokens are valid and have not expired. I am wondering if there is a specific rate limit or capacity constraint for the analytics API that is being hit during high concurrency, or if this is a known issue with the WebRTC SDK when used in a load testing scenario. Any insights into how to handle these 503 errors or if there is a recommended pattern for fetching analytics data during peak load would be greatly appreciated. I have attached the JMeter thread group configuration details below for reference. The key settings are: Loop Count = Infinite, Ramp-Up Period = 60 seconds, and Scheduler = On with a duration of 300 seconds. I am also using the Standard Timer with a constant delay of 1000 ms to space out the requests.