[nabilhat]

In this case it's a bad test thing.

The car impacted at a barrier splice which was not correctly joined. Barrier splices can come apart in the real world, but the fasteners will tear apart the sheet metal in the process. That splice gently deflected apart, like two sections of barrier were gently leaned against each other.

Two posts fall over on the initial impact, one on each side of the loose splice, without causing noticable specific damage to the car. The third post absolutely caved in the front end of the car before being violently bent down underneath, as properly anchored posts do in the real world.

This also isn't how this type of barrier is tested. "Barrier deflections listed below are results from crash tests with a 2,000 kg (4,400 lb) pickup truck traveling 100 km/h (62 mph), colliding with the rail at a 25-degree angle." More diverse testing done with seriousness would be a useful way to challenge if that standard continues to be appropriate, but this isn't it.

https://en.wikipedia.org/wiki/Traffic_barrier#Stiffness