Fab4's report of the number of milliseconds played is not very stable with respect to the associated reported timestamps. This means SC needs to use the multi-packet averaging mechanism to determine a play-point which can be used for synchronization calculations, which in turn means that resynchronization at the start of playback is typically delayed until about 6 seconds into the track rather than almost instantly.

I suspect that ALSA drivers are the cause of the problem here and we should look to see if it is possible to improve the reporting. The current variation is of the order of 8ms between reports.