The problem with those presentations is that they rely on a future in which the other cars are driving in a highly structured way. I.e., they are also self-driving cars.
This car has to be 100% certain it will not be the cause of an accident. It is likely tracking the speed of the vehicles around it, and if they aren’t consistent enough and spaced far enough for it to achieve a given confidence level, it will just stay put until a large enough gap opens, or until the traffic stops in the next lane.
Personally, if I recognized a Waymo vehicle, I would stop for it to allow it to merge. But obviously that is not ideal for the flow of traffic.
It’s for this reason that i think the way self-driving vehicles will achieve saturation will be for drive assist technology to continue to improve and become more standardized. That will take a long time, probably at least another 10 years. Then, it will become more likely that Self-driving cars can merge effectively and fluidly.
I feel like what might need to happen is for there to be some sort of external signalling showing when cars are in an “autopilot” mode, like blinking blue lights on their front and back. That might enable other autopiloting cars to recognize and synchronize effectively. But, who knows. I’m certain there are problems implementing that sort of standard.