Good question. We don't delay signals since the user may walk between rooms.
Syncing is an interesting problem for sure. It's really a question of how precisely you can sync the clocks between devices. Our current method is fairly naive, we let ntp sync clocks for us, count the number of samples played, and drop samples every now and then on devices that are behind. This gets us to 2ms accuracy between devices. We could get much lower with a better algorithm. But 2ms is very difficult to notice in a home listening situation between rooms.