If the videos you refer to are , I'm not entirely convinced. I have a suspicion that you, and the poster of the videos, think that the correct behaviour for someone restrained in a road car is for them to remain fixed in their seat during an accident. The reality is that in a high-energy accident (and that looks like a high-energy accident slowed down, although it's a bit naughty to have no scale on the images) you want the accident to go on for as long as possible, and if that's at the expense of moving forward under the control of the belt, so much the better.
You don't die of your arms and legs flailing around. You do die of your neck being broken (which is why rear facing seats are absolutely essential for small children) and your internal organs and brain banging against your ribs and your skull. You reduce that risk by stopping as slowly as possible, as far forward as possible without hitting anything hard. The stretch and movement in the belts, which are often interpreted as "bad", are in fact what's saving your life. An infinitely strong seatbelt which held you exactly in place would be far more dangerous, because rather than having the collapsing of the crumple zones and the stretch in the belts, you would stop stone dead (figuratively and, given some speed, literally) as soon as the crumple zones were used up.
This stuff has been measured, in detail, as part of competition car homologation. The regulations now give maximum g-values for the chest and head areas, which require very careful engineering of the belts to provide enough stretch (Mika Hakkinen hit his head on the wheel in the 1990s, but had he not done so, he would have taken a massive g force and probably died). Competition drivers are now mostly wearing HANS devices to stop their heads from moving, so the need to protect their necks is less, but there is a lot of a dynamic movement in the seatbelts to prevent internal injuries. Simply watching crash-test images isn't going to help if you're being upset by limbs flying around: that's not the key issue.