Seriously with that many units, why not zerg from the beginning?

short answer: No one wants to be the first one to run out, and be assured a beating. And there's no way for everyone to decide to run together immediately.

long answer:

Obviously no one in the mob/crowd has much more power than anyone else in it. That means that there's no method of the group to signal to itself to rush at any one moment. Obviously everyone in that crowd wanted to do so, and knew it was coming (tis why they never stopped approaching the cops), but the individual(s) that runs out too far ahead of the pack will be beaten horribly by the police before the mob gets there. Even worse, the crowd having to watch that will likely sap the crowd's motivation.

So what's needed is for the crowd to, in a moment of group think, all decide that 'this is the moment,' and rush in together. This happens in steps. In the middle of the crowd, people are going to respond and follow to whatever the closest people to them do, so the initially important people are in the front.

The people in the very front are the ones trying to initiate the group think moment. You can see they each run out ahead really quickly, throw a stone or remove the tear gas, check to see if they're being followed yet, and if not, retreat and repeat.

Since these people are running out at random, eventually enough of them will run out together, or enough of the line of people will happen to join them, in close enough concentration, that a critical point will be reached. This will be when so many people in one spot start moving, that the people around them figure "this is it," and will start going, and it will start with enough people in the initial moment that the entire pack will get the message before anyone has a chance to hesitate and back down.

This also means two things. 1) You can see tons of false starts on every level before the critical point. Individuals starting and restarting, and small groups starting and restarting, and even at one point, the entire left half of the crowd starts running, but not enough people jumped forward quickly enough to be at the critical point. 2) It explains why the riot police, with a few stones and especially tear gas, are so effective. Stone throwing, and especially tear gas clouds make it much more difficult to be one of those people in the very front trying to initiate the crowd. If the riot police can hamper these people enough, the crowd will be less likely to rush. So those guys throwing back the tear gas are incredibly important. And brave too, because if they show themselves being too hurt, then fear could sweep through the crowd. So they have to be successful.

This is also why we elect leaders.