Trying to understand why predictive routing scores for digital channels drop significantly when a bulk export job is active. We are running a legal discovery export via /api/v2/recording/bulkexports on Genesys Cloud (Europe/London region) and notice that the WEM analytics data for concurrent chat sessions shows inconsistent routing confidence metrics. The export job uses standard scopes but seems to lock metadata resources, causing the predictive model to revert to historical averages instead of real-time sentiment analysis. Is there a known resource contention issue between bulk export processes and the predictive routing engine for digital media types?