Redundancy sends every voice packet again, one to four times, 20 to
200 ms apart. Listeners' jitter buffers grow to wait for the copies,
native clients' included, so a lost packet costs nothing while a copy
arrives. With a native client listening over a bursty lossy uplink,
the frames lost to concealment fell from 10% to under 1%.
Android gets the desktop's encoder settings: bitrate, complexity, VBR,
FEC, music codec, and the new redundancy delay and copies.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>