This bug has been found and fixed in test releases.
As for the latency, what I will say is that IP based multiroom synchronised audio, such as used in Volumio, makes very different design decisions to those in professional sound engineering. The latency and jitter of your home network stack is already a couple of orders of magnitude bigger than 0.1ms (more again if you are on WiFi). To run at acceptable CPU load the Linux sound stack usually a buffer of at least a couple of hundred ms. To get acceptable network performance you need to send approx 5ms of CD audio per packet.
The snapserver buffer is therefore sized to be big enough that data can be reliably and efficiently propagated to remote clients, and fill their buffers, with some headroom for if the network gets choppy.
Could the buffer be smaller? If your network is all gigabit wired, and you’re on Pi 3 or 4 level hw, then yes. absolutely. Would that be a sensible default? Probably not. Many volumio users use WiFi, and a lot have low power pi 0 devices in the mix. 1500 ms gives a good experience out of the box, even on a fairly poor network.