At my company, we have around 400,000 servers in production. When we last surveyed them, we found several thousand over 12 years old, with the oldest at 17 years. And that wasn’t counting our lab and admin servers which could run even older because they’re often repurposed from prod decomms.
We had a huge internal effort to virtualize their loads, but in the end, only about 15% were transferred just due to the sheer number of hidden edge cases that kept turning up.
I think they’re trying to simplify the exposed interfaces simplifying everyone else’s job at the expense of making a more complex implementation.