I don't find it surprising at all, sadly. After a reorg, one of my colleagues on a different team (largely Master's degrees in ML) told our new team member: 'You shouldn't be worried, being assigned to their team is an honor. They're hardest to get into -- it's full of people who didn't just skip all the hard courses.'

In many ways, hiring advanced degree holders is a crapshoot. They have skills you probably can't train, but often times come in with fewer software development skills than your undergrad intern, despite theoretically having more years of experience. You don't need to know git to publish in IEEE, or write unit tests or readable code, or debug an edge case, and your only code reviewer is a professor who doesn't care about this either. 'Good enough to publish' is a far cry from 'customers will pay for it.'

I don't think it's fair to characterize such people as "incapable of learning how to do new things" if they're not even being given the opportunity to learn.

I'm not saying that the OP should hire these people, or that they would even be good hires. Maybe they don't want to take the risk or time of training up a new employee.

It just seems like a bit wierd non-sequitur to draw conclusions on someone's learning capabilities based on their performance in an interview outside their usual domain.

I never said they were incapable of learning new things. But they're usually not applying for junior level roles where we expect to make these investments. They are typically applying for senior roles where the applicant is expected to possess these soft skills.