We do use variable bit rate (VBR) vs. constant bit rate (CBR). The bit rate is dependent on input speech and adapts to significant network changes.

1. Yes, there is a theoretical possibility that somebody could extract information from the encrypted data stream.
2. As stated in the IETF document (the topic itself was highly controversial within the IETF), VBR only is a problem for short pre-recorded or repetitive speech snippets.
3. The data leak due to VBR has been shown on a theoretical level with a limited vocabulary. So far, to our knowledge, there is no known attack against naturally spoken speech.
4. We don’t stream pre-recorded audio messages, they are sent as binary files.

Thanks for pointing this out to us, we will consider an even more secure option in the future.