This is a read-only counter, in milliseconds, tracking how much voice playout was synthesized from redundancy parameters available to the decoder, due to voice data arriving late or being lost, and like the other gap-fill counters it locks at roughly two days. An admin monitors this to see how much the call relied specifically on redundant data, rather than prediction or silence, to mask lost packets, indicating the peer's redundancy feature is actively compensating for loss. For example, a high redundancy gap-fill value would tell an admin that packet-loss recovery via redundant data is doing significant work on this call, suggesting real underlying network loss even though the call sounds acceptable.