Realtime API - Message being cutoff followed by silence

Hi,

We have been using the Realtime API since its preview release and have noticed an issue where the model suddenly stops responding for several seconds before sometimes resuming.

In some cases, we observed pauses lasting as long as 30 seconds.

We’re using the newest model version - gpt-4o-realtime-preview-2024-12-17

Is this a known issue, and are there plans to address it in the near future?

By the way, we have the transcript, recording and session id - in case you need it for debugging.