Has anyone figured out why the /api/v2/predictiverouting/queues endpoint reports inflated queue depths specifically for interactions routed through our Singapore BYOC trunks? The SIP registration logs show stable connectivity, yet the predictive analytics engine seems to miscalculate available capacity during peak APAC hours. This discrepancy causes unnecessary agent idle time while calls queue artificially. The issue persists across all 15 trunks despite identical routing logic to PSTN routes.