In the second, the vehicle was allowed to drive across a traffic lane. And in fact it attempted to do so, but stopped at the threshold when it saw a vehicle approaching (which also hit the brakes). I fail so see how this isn't exactly the behavior claimed and desired.
If driving into live traffic and stopping in middle of it is acceptable, then okay, self driving cars are here!
It pulls out as if it were going to continue across the traffic lane and very clearly stops right at the threshold. At no point is it in the traffic lane, nor in the path of the gray vehicle that stops.
It's true that it "looked like" a human driver who was going to continue straight out, so the other driver (correctly) hit the brakes.
But spinning this like the Tesla "stopped in the middle of traffic" is just plain wrong.
And of course it hadn't "slowed", at this speed (looks like about ~1 m/s) the time taken to stop (at about the .7G a car on its wheels can achieve) is about 4 frames of that 30 Hz video, and it happens over a distance of about 7cm, most of which is the vehicle just rocking forward on its suspension. Be real, we're talking about parking lot maneuvering here.
Okay, now we need to do some math. The evidence that a 4,000lb vehicle coming to a stop even from 5mph in... 3 inches? "Most of which is the car rockingon its suspension". Removing distance from the equation, because g forces are related to time of deceleration, using your numbers, https://rechneronline.de/g-acceleration/ says more like -2.3G which is far from a gentle stop.
You can also plainly see the other car decelerate too, over far more than '4 frames'.
I suppose that doesn't really fit the narrative you're trying to spin here.