Currently PulseAudio does a pretty good job of figuring out the network latency and reporting it back to the applications. But this is probably not what we want.
The main reason applications wants latency information is to synchronise audio and video. But this is based on the assumption that video latency is negligible. But for ThinLinc, the video will suffer the same network latency as the audio.
So in order to have good synchronisation over high-latency links, we should lie to the applications and say that the network latency is non-existent.
This will possibly brake some other use cases, but I think those are so incredibly rare that this primary use case is more important.
PulseAudio 3.0 has "Configurable device latency offset". Perhaps that helps?
This delay is calculated by the clients, so we cannot easily adjust it from the server. After doing bug 4194 it will be trivial though.