It isn't at all.
PIP is a gateway benefit, no PIP, no access to other things for example, in some areas, no PIP, no blue badge. No PIP, no carers allowance, no way of getting a carer into various places. No PIP, no motability vehicle.
Truly millionaires on the whole, don't bother jumping through the PIP hoops if they don't need the above, because it is so hard. There are few of them, realistically speaking.
Generally 'well off' people do, because of the above, those who need to prove some sort of point (ie, 'the system works') do (politicians!)..
Means testing adds an extra layer of paperwork and expense - it is why heating allowances for pensioners are not means tested, it would cost more to do it than just giving them all the money.
Government do like a way of generally counting the disabled people and now we do not hold a register of all the cripples in the country, this is the next best thing, means testing would alter that significantly.
The few outliers who truly are so astronomically wealthy that the costs of being disabled are irrelevant, are so few, and the numbers of people who would wobble between 'entitled' and 'not entitled' as savings/income waivered around the cut off point would be so huge, it would be a total fucking nightmare to manage.
How would you, would you hold a database of people who are disabled but not entitled, and just grant them the benefit when savings dip, how long would savings have to dip, would you be giving them a motability car for six months then taking it away when Uncle Fred kicks the bucket and leaves them 10K?
Or would you re-test from scratch, in case their disability got better in the time their funds ran out? It is already taking months to YEARS longer to sort out claims than it should...
To prevent a tiny tiny number of people from getting a benefit they financially don't need, you would be denying a vast number of people who ARE in need.
We're already doing that for other reasons and it is causing shocking suffering, I don't really think we need to do it more (reducing successful claimants by 20%, when the fraud figures are less than 1%...)