The video does not comment on the situation which arises when a car stops for a prolonged period, blocking the cars behind it. What would or should happen in this scenario?
Given that the cars can't overtake each other, and that it's unthinkable to move a car when its safety door is blocked open, then inevitably the pair of cars behind the stopped pair will have to reverse direction until they come to the first available exit/entrance and stop there with doors open until the blocked safety door is cleared and the original stopped pair can move again. Obviously a parked car-pair will also impede the car-pair behind that, so in fact the whole system will have to come to a stop. But the operation of a conventional elevator shaft is likewise completely blocked for as long as the safety doors are blocked open: probably the major difference is that the result is less waiting inside a blocked elevator and more waiting for a blocked elevator. A normal two-elevator system will at least have one elevator working as normal while the other is stopped by a long-blocked car, but OTOH short blockages should delay the Hitachi paternoster system less thanks to its element of parallelism. Also, long door blockages could get sticky on a Hitachi paternoster without a stop on every floor, but I assume that the intended common-case solution for this is "don't do that, then". (For similar reasons, implementing restricted-access floors on a Hitachi paternoster would be problematic.)
I'd assume that the cars stack up, and then a following car skips a stop in order to re-distribute, like a bus or subway line. The an elevator is a more concentrated system with more easily collected data, so whatever algorithm is controlling it would more easily detect patterns. So the stacked cars might not skip, in the case of say, a horde of people heading from the ground-floor of a hotel to a keynote on the ballroom floor, but might skip in the case of an anomalous delay like someone with lots of packages.
