WebRTC Softphone Audio Latency Spikes in Bulk Export Metadata

How come this setting causes significant audio latency spikes when exporting WebRTC softphone interactions via the bulk API? We are currently processing a large volume of legal discovery requests for digital channel interactions, specifically WebRTC softphone sessions. The environment is Genesys Cloud EU-West-1, and we are using the POST /api/v2/interactions/bulk endpoint to initiate export jobs. The issue manifests when the interaction metadata contains specific WebRTC diagnostic fields, particularly the remoteCandidate and localCandidate SDP details. When these fields are present and populated with extensive ICE candidate lists, the resulting media files in our S3 bucket exhibit a 200-500ms audio-video desynchronization. This is not a playback issue within the Genesys Cloud UI, but a corruption in the exported WAV/MP4 files themselves. The bulk export job completes successfully with a status of COMPLETED, but the chain of custody audit trail shows a mismatch in the media duration versus the interaction duration recorded in the interaction archive. We have verified that the S3 bucket permissions and encryption settings are correct, and the issue persists across multiple export jobs initiated within the last 48 hours. The problem does not occur with standard SIP trunk interactions or PSTN calls, only with WebRTC softphone sessions where the client browser is Chrome version 114 or higher. We suspect the metadata processing pipeline for WebRTC diagnostics is interfering with the media stitching process during the bulk export. Has anyone encountered similar media corruption issues related to WebRTC SDP metadata in bulk exports? We need to ensure the integrity of the exported media for legal compliance, and any desynchronization compromises the admissibility of the recordings. We are considering filtering out the WebRTC diagnostic metadata before export, but this may impact our ability to troubleshoot future network issues. Guidance on how to handle this metadata without compromising media integrity would be appreciated.