My current config is completely failing… We are using Genesys Cloud WebRTC softphones for agent verification of bulk export metadata. Audio drops every 45 seconds during high-load periods in London. Network traces show consistent jitter but no packet loss. The Architect flow routes correctly, yet the client-side SDK v5.2 logs ICE_CONNECTION_FAILED intermittently. Has anyone tuned STUN/TURN settings for legal hold review environments?
Thanks for the help.
What’s probably happening here is that the default stun server timeout combined with the aggressive nat traversal settings in the sdk v5.2. in london, we’ve seen similar jitter spikes during peak hours. try overriding the ice transport policy in your web client config. set it to relay only if your network blocks udp, but usually all with a custom turn server works best. here is a snippet for your data action or client init:
{
"iceServers": [
{
"urls": ["turn:your-turn-server.eu-west-2.amazonaws.com:443?transport=tcp"],
"username": "creds",
"credential": "secret"
}
],
"iceTransportPolicy": "relay"
}
also, check if the bulk export process is hogging bandwidth on the agent’s network interface. the sdk logs showing ice_connection_failed often point to a firewall dropping keep-alive packets. ensure your turn server supports tcp fallback. this usually stabilizes the connection during heavy load.
The best way to fix this is…
Cause: SDK v5.2 default ICE timeout conflicts with London region latency spikes.
Solution: Override iceTransportPolicy to relay and increase iceCandidatePoolSize to 10 in client init.