There’s a lot, and specifically a lot of machine learning talk and features in the 1.5 release of Opus - the free and open audio codec.
Audible and continuous (albeit jittery) talk on 90% packet loss is crazy.
Section WebRTC Integration → Samples has an example where you can test out the 90 % packet loss audio.
I wonder if/when this will be available in voice chat clients like TeamSpeak.
It requires compile flag and runtime enabling. So it at least requires some thought and development more than just a lib version upgrade.
I saw the link on the Mumble dev chat, so at least they’re aware of it.
Given that it’s only really (very) useful in very flaky/bad connection scenarios I’m not sure it’s even worth it in most cases.