But is that based on research performed, or just a "better safe than sorry" statement? Have they (or anyone) ever seen a seat fail prematurely or perform worse due to being involved in multiple accidents? That page doesn't have any more information other than that statement. I'm having a difficult time imagining what parts of a car seat would be damaged during a crash (where the seat itself is not impacted). One exception is if car seats are designed to deform, crush, etc. to absorb energy, like a bike helmet is. But I've never seen anything like that.

Also, when trying to decide whether a second-hand car seat is safe, you're multiplying the probability of a seat being damaged in an accident, by the probability that the damage will significantly reduce the safety of the seat, by the probability that the seat being sold was in an accident. I understand if a parent would rather just buy a new one and not worry about this. But I see this as being a near-zero risk.

edit: I see that some child car seats do use EPP foam, which could possibly get permanently compressed during a severe impact.

Fair enough. I buy it more for bike helmets, though, since there's a good chance that helmet hit road or metal.

And in an accident a car seat has all the force transferred through just 2-3 points of contact with the seat belt or isofix points.

I would replace my kids seats after any accident with airbags or pretensioners firing.

The damage on a bike helmet that has hit the road or metal in an accident will be very obvious in even the most cursory visual inspection.