Odd how my premiums have been a decent level over 70 until now.
I think most people's premiums went up this year due to inflation, and some of the other points you mentioned around parts availability. How much was that, and how much is anything to do with getting older is probably not possible for any of us to say. Until a couple of years ago I was working with insurance companies' risk software, and I know a little bit, but exactly how that software is used is a very closely guarded secret - the art is in creating a quote that will cover payouts but be commercially enticing to customers and leave some profit. Even if I knew the detail I would have had to have signed an NDA, and certainly wouldn't be telling anyone, especialy with my name on the post! Maybe an anonymous data scientist in the group would care to share
Younger age groups have higher premiums too because of more expensive claims, not just older, and claims to insurance companies in £££ get spread across the board to an extent to ensure their massive profits, so certain sectors/ages seem to help them achieve this rather than based on an individual drivers history.
Not sure how massive the profits are in motor insurance. They always swore it wasn't much when it came to renewing software licenses

And people who sell software licenses definitely read company accounts to understand how much money might be floating around.
There may be some aggregating of policy costs across age groups (the way people is grouped is done on a lot more factors than just age, but I think age is significant), but I have never seen this mentioned in risk calculation. It may be something that was getting applied after any part I had in the process. I don't think it is significant. You may know more?
Forecasting the probability of future events doesn't really work if you only observe one actor. Say we have 5 years of data about you and your crash record is 0,0,0,0,0. There's lots of ways we can estimate year 6, but the very simplest is the mean. Year 6 is (0+0+0+0+0)/5 = 0. So I can estimate you have 0% risk of crashing next year.
But I have this other driver, same age, same location, and their last five years is 0,0,0,0,1. Their mean is (0+0+0+0+1)/5 = 0.2. So I can estimate they have a 20% risk of crashing next year.
But also what I know is that just because you have no crashes for the last 4 years does not mean you won't ever have a crash. So I need to load some of this risk onto you. And the risk engines are clustering customers on lots and lots of factors, and aggregating the risks among customers who are members of the same cluster. But not, as far as I am aware, over different clusters (at least significantly).
Also, in reality no one estimates risk on mean. They use something like (but generally more sophisticated than) exponential smoothing where I care increasingly less about how you drive the longer ago it was, because most forecatsing models agree that the longer ago an event occured the less influence it has on what will happen next. It's why NCB makes a big (but decreasing) difference in the first 1-4 years, but then very little more.
TLDR; your individual driving history can only be a small part of any risk calculation.
In the older sector, based on your findings, it seems that a medical or 3 yearly license imposed at certain ages, means /does nothing to keep those that pose a greater risk off the road regardless, and the general standard of driving/accidents by many twats in the 'in between ages' costs everyone incuding you and me.
Those numbers are costs per claim. The medical and license restrictions may mean there are fewer risky people on the road. But those who are on the road and do have accidents cause larger claims. My guess would be there is more of a risk of injuring people which is generally what costs a lot of money. Much more than scraping your Bentley, for example.
Its not just those who have accidents that are penalised, but that's how they work their system, higher premiums for accident prone and higher premiums for the rest as a result of it.
If it was just those who had accidents who were "penalised" there would be no insurance. You'd just pay if you had an accident. Insurance is just aggregating your risk with others, and the way risk engines are built those others are people with similar characteristics to you, be that age, location, driving history, vehicle type, etc.
Lot of waffle really, but bottom line is that my understanding is that there is very little influence on your premium from idiot drivers 20+ years younger than you. It's not commercially viable for insurance companies to do that. It would only take one to disassociate the clusters and they would win a load of low risk business. Unless of course it's all a big price-fixing cabal *dons foil beret*.