All the bad outcomes that you’ve highlighted about not putting users at the center of research are even more true when you are specifically thinking about users with disabilities. None of your personas have disabilities? Well, 18 % of the global population does. Why not give Jason, the cloud admin, some hearing loss from keeping those headphones on too long at too high of a volume when he was a teenager? Give another user carpal tunnel (so no mouse) and someone a temporary skiing injury.
Working with people with real disabilities when doing user research can be quite the revelation. You can have two people with hearing loss and identical audiograms and they adapt completely differently, but you don’t know that if you don’t include them