Queue observation API returning stale data during peak load

I’m hitting a wall with the Genesys Cloud Statistics API. We need real-time queue metrics for a custom dashboard, specifically the number of waiting interactions and available agents. The documentation points to GET /api/v2/analytics/queues/realtime/summary with interval=PT0S for point-in-time data. It works fine in staging with low traffic. In production, during our morning rush, the data is consistently 2-3 minutes behind the actual queue state. I’ve checked the granularity and interval parameters. I’m using interval=PT0S and granularity=PT5M just to be safe, but the lag persists. The response headers don’t indicate any caching issues. I’ve tried adding forceRefresh=true as a query param, but it gets ignored or rejected with a 400. Here’s the request and response snippet:

GET /api/v2/analytics/queues/realtime/summary?interval=PT0S&granularity=PT5M&forceRefresh=true
Authorization: Bearer eyJhbGciOiJSUzI1NiIsInR5cCI6IkpXVCJ9...

Response: 200 OK
{
 "data": [
 {
 "queueId": "12345",
 "queueName": "Support-DE",
 "metrics": {
 "waitingCount": {
 "value": 5,
 "timestamp": "2023-10-27T08:15:00.000Z"
 },
 "agentsAvailable": {
 "value": 2,
 "timestamp": "2023-10-27T08:15:00.000Z"
 }
 }
 }
 ]
}

The timestamp in the payload is always 2-3 minutes old. The actual UI shows 15 waiting. Is there a different endpoint for true real-time data? Or is there a specific header I’m missing? The forceRefresh param doesn’t seem to do anything. I’ve checked the API version. We’re on the latest stable. No luck.