Can anyone clarify the impact of active recording bulk export jobs on predictive routing calculations for digital channels?
We are observing a significant drop in routing scores for WhatsApp and Web Chat interactions when a large-scale legal discovery export is running. The exports are triggered via the /api/v2/recording/bulkexports endpoint and target our BYOC S3 bucket in eu-west-2.
The issue started after we increased the concurrency of the export jobs to meet a tight deadline for a chain-of-custody audit. When the export job status is ‘running’, the predictive routing engine seems to deprioritize agents who are currently on digital channel breaks or have low historical interaction volume for those specific channels. This results in increased wait times and missed SLA targets for high-priority customers.
We have checked the system health dashboard, and CPU usage on the routing nodes is within normal limits. The metadata in the exported files matches the original recording properties, so data integrity is not the issue. However, the routing logic appears to be influenced by the backend load generated by the export process.
Is there a known throttling mechanism or resource contention issue between the recording export service and the predictive routing engine? We need to ensure that legal holds do not negatively impact operational performance during peak hours.