The TURN server works just fine when used on a single-home system with default IP settings, or when used on multi-home systems with explicitly defined listener IP(s) and relay IP(s). But what about the case when the TURN server is running on a multi-home system with default settings (no listener and relay IPs are explicitly defined) ? Currently it will not work well - all IP addresses will be used for the relay endpoints, and as the result some relay endpoints will not be reachable from some networks.

The default policy must be "use the same IP for the relay IP address as the local listener IP address for the connection is". This policy will work perfectly for public TURN servers. This default policy will NOT work for "gateway"-type private and semi-private TURN servers with multiple network interfaces, but those TURN servers must be set "manually" and "explicitly" with the listener and relay IP addresses, anyway. For those "private" TURN servers nothing will change - they still have to be set explicitly and they cannot use any default policy anyway (because the TURN server cannot determine the administrator intentions in complex network topology and it has to be told what to do).

It will be implemented in 3.2.2.2 version

Fixed and will be released as part of 3.2.2.2 build.