Sendspin’s Python server corrects timing failures that could leave grouped speakers audibly out of step after audio production stalled. Different formats could end up 15 to 357 milliseconds apart, and pausing and restarting did not always realign them.
Maxim Raznatovski makes the audio paths follow the same timing changes and closes gaps when a speaker catches up to an ongoing stream. Music Assistant also corrects a separate handover that could leave a speaker using equalizer filters about 100 milliseconds behind when it joined a group.