That's a little different. While Jamulus tries it's best to do realtime audio, ninjam works deferred.
Players agrees on a bpm and a number of bars, the software will record each bar and send to other that will cache them and then play them on the next loop.. so every player will hear others one phrase late.
(sorry for non precise terminology...)
https://gist.github.com/fabrixxm/5c072ac64b680cb4a3406e2763d...