https://www.newyorker.com/magazine/2011/04/25/the-possibilia...
In high school physics we were doing some experiments around pendulums and used metronomes to measure the period; I remember the teacher telling us about a former student of hers, a concert level musician, who could keep a beat in perfect timing without using a metronome at all.
And yes, digital audio typically inherently has some latency due processing typically being based on some buffers. Sure, you can try to make those buffers smaller but that is not always so easy to do.
On a full duplex phone call, you can tolerate quite a bit of latency, upwards of 600ms, provided there is no echo, as soon as echo is present, we notice the latency - but even on echo free calls, the longer the latency, the more likely both parties will end up speaking at the same time.