A dramatic increase in amount and severity of accidents after a major world-impacting event like the pandemic seems like exactly the sort of thing that would make sense to be some change in the way people drive. This isn't "Oh, the users just hate the new Slack design. What idiots." For 90+ percent of the US, nothing has changed about the cars or roads in the last 4 years any more than the 4 years before that, and yet the outcomes of driving have changed.
Bear in mind that the auto insurance industry is quite distinct from the auto manufacturing industry.
Due to increasingly pedestrian-unfriendly designs, like "grr I'm a bad boy" grills on trucks, crashes becomes more damaging to what they crash into. More, previously non-severe crashes becomes severe crashes.
The purported timeline is: Pandemic hits, people stop driving, incidents falls drastically. Pandemic stops, people starts driving again, but incidents and severity increase drastically compared to pre pandemic.
The companies are saying internally "people forgot how to drive properly", which granted, sounds weird. But it also sounds weird that people upgraded their cars, while not driving them, or that there was a huge spike in car sales as people upgraded post pandemic.
Sales data doesn't suggest that consumers went out and changed their fleet [1]
Well, first of all it doesn't need to be what any person is convinced is "large", it just needs to be nonzero. Second of all, it's not about customers, it's about manufacturers. As your own data indicates[0], a nonzero number of people replaced their car during that time period, just like in every other time period. If more people choose electric than did decades ago, there will be more electric cars than decades ago. And there are[1]! Electrification is increasing, therefore weight is increasing, therefore crash severity is increasing.
So the timeline is, each year, more and more people are buying electric cars, and it looks like severity increased, too. Do we have data on people's ability to drive over time? Maybe, but I haven't seen it, so that's what seems unconvincing so far. We definitely have data on vehicle weight over time, though.
[0]: "Sales data doesn't suggest that consumers went out and changed their fleet" – by my viewing of the data, it shows that a lot of consumers, in fact, went out and bought a car during that period. Millions did, tens of millions even, and it only takes a single car difference to affect the population average.
[1]: https://www.bls.gov/opub/btn/volume-12/charging-into-the-fut...