Comment by Animats

I briefly worked on synchronized applause (well, a toy discretised cellular automata-like model of such), and the individual agents don't even need to know the mean, or to receive continuous feedback (which is what happens with synchronised metronomes).

As long as they can infer a collective pace from listening to the general loudness in the room, they can do very basic adjustments to their own clapping rhythm=phase and they will get in sync.

And applause is just a stand-in for any locally periodic behaviour with regular signals (claps) which can be aggregated (more claps = louder).