Aggregation logic failing on tHandle/tAcw for 30-minute intervals in Analytics API

We are attempting to build a granular agent utilization dashboard that breaks down handle time, after-call work, and hold time into 30-minute buckets. The goal is to feed this data into our Terraform-driven reporting infrastructure, which requires precise JSON structures for downstream processing. The endpoint in question is POST /api/v2/analytics/conversations/details/query.

The issue arises when we adjust the granularity parameter. If we set granularity to 86400000 (one day), the response includes the expected metric objects with correct values for tHandle, tAcw, and tHold. However, as soon as we switch to 1800000 (30 minutes), the metric arrays in the response body become empty or return null values for the time-based metrics, even though the conversation count remains accurate. We have verified that the date range covers active working hours and that the group filter correctly targets the support agents.

Here is the relevant portion of our request payload:

{
 "dateFrom": "2023-10-23T00:00:00.000Z",
 "dateTo": "2023-10-23T23:59:59.999Z",
 "granularity": "PT30M",
 "metrics": [
 "tHandle",
 "tAcw",
 "tHold",
 "conversationCount"
 ],
 "groupings": [
 {
 "type": "wrapupCode"
 }
 ],
 "filters": {
 "group": {
 "type": "in",
 "values": ["support-agent-group-id"]
 }
 }
}

The API returns a 200 OK status, so there is no server-side error. The response structure is valid JSON, but the data points for the time metrics are missing. We suspect this might be related to how the aggregation engine handles sparse data for shorter intervals, or perhaps there is a specific metric type required for interval-based queries that differs from daily totals. We have tried switching the metric keys to their string equivalents found in the documentation, but the result remains the same. Is there a specific configuration or additional parameter required to populate these metrics for PT30M granularity? We need the raw numbers to calculate utilization percentages without relying on pre-built reports.