I had a look into how Jamulus works under the hood. It’s more of a hub and spoke model with the central server receiving, buffering and resending the stream out to each client. This has some advantages (kinder to bandwidth especially as the number of clients increases) but it does mean that there are two sets of recv buffers (server and client) which need to be deep enough to avoid too many missed packets due to jitter. My personal experience was that even when running it all on a lan I couldn’t get much below 16-20ms with an ideal near zero latency network. Once you add in geographic latency and internet jitter then it goes up considerably.

(Full disclosure: Weepy and I played with Jamulus then he had a crack at building sub.live)