Looking for advice on persistent WebRTC softphone disconnects occurring during peak hours in the Europe/Paris region. The Architect flow logs show successful connection events, yet the queue_activity view indicates immediate abandonment with no 504 or 502 errors in the backend logs. The issue appears isolated to specific agent_performance metrics where talk_time is recorded as zero despite caller retention. Is this a known limitation with the current softphone SDK version or a configuration gap in the conversation_detail routing logic?
Make sure you verify the WebRTC signaling timeout configuration in the Architect flow, as the default values might be too aggressive for high-latency connections in the Europe/Paris region. While the logs show successful connection events, the immediate abandonment suggests the media stream fails to establish before the softphone drops the connection. This often happens when the browser’s ICE candidate gathering is blocked or delayed by firewall rules.
Check if the agents are using corporate networks with strict NAT traversal policies. The talk_time being zero confirms no media packets were exchanged, which points to a media path issue rather than a signaling problem. You should also review the browser console logs for any WebSocket or ICE errors.
- WebRTC ICE candidate gathering
- Firewall UDP port restrictions
- Architect flow timeout settings
- Browser console media errors