Best I can tell the coordination happens with a half-dozen to a dozen people mixed into the middle and back of the crowd. Something similar to
how it worked in Chicago, but without the obvious ponchos.
They seem to be operating on a "know their face" system, to prevent infiltration; the core group knows everyone inside, and if someone shouts directions contrary to them, they are quick to call them "outsiders". The larger mob knows at least 1 of the core group. It's a crude
mesh node network in human form.
The problem now is that ever since that one article about the feds using streams to identify people, streamers are
not recording the crowd. They are intentionally filming only the cops and the front of the protest line. For all the shit we (rightfully) give them about LARPing, they understand social media and 4chan autism, and they have
some OpSec going on in this area.
Whenever you see one of Antifa's designated rioters throw the big stuff like fireworks or mortars, they always approach from the back of the crowd, throw from the middle, then retreat again. The guy
handing out light bulb paint bombs was also in the back.
I think the megaphone people can either be organizers, or useful idiot distractions. Someone loud and obnoxious in the front can keep all eyes and cameras on them, so you don't see the communication in the background. They don't even have to be part of the inner circle, just give them the megaphone and route directions and let them do their thing.
It also fills their narcissistic need for attention and a feeling of power, and lets the organizers control optics. Everyone who says "3 black persons at the front and 99% white crowd" is noticing something they aren't supposed to notice. But you may
also notice the megaphone bitches are
conveniently never in the front of the crowd when the tear gas goes off, or the police begin their charge. They're always in the middle or the back, so they can re-rally the morons after someone gets stomped by the cops.