That suggests one of three things:

Amazon had a glitch.
Amazon was experimenting with something, and the experiment didn't work out as planned.
Amazon listened to feedback and adjusted accordingly.

I'd like to think it was the third, but we'll never know.

The other day, I had a self-check out machine give me attitude. Someone actually recorded (or AI'ed) a condescending tone and thought it was a good idea. I swear to God this isn't a joke.

I believe you.  Idiocracy is rapidly becoming the most prescient dystopian story ever written.

Fixing the problem cuts into resources, too.

Further, do they care?

Amazon's going to Amazon, you know?

Giving a non-vague error message leads to a swifter resolution, perhaps without consuming customer support bandwidth.

After all, they have to deal with ornery people like me who are more than happy to keep pushing a question like that until I get an answer. Even if a lot of the "people" in the chain are bots, all of them aren't. So their resources get consumed offering support on questions they could conceivably answer with the initial message.

This is not necessarily a small issue. I can and have cut all ties with businesses that couldn't be bothered to have good customer support. I'm sure I'm not the only one. Amazon may see indie authors (who are, after all, customers of the KDP service) are a captive audience, but you yourself can clearly see a potential non-Amazon path, or at least a non-exclusive one. It would take far less effort to address some of these simple issues with clear error messages and info rather than following the vague-error-message path.

Lots of tech companies give vague or even insulting error messages.

A long time ago, I had a webpage shut down by one of these companies for a non-specific rules violation.  The error message that displayed to visitors to the page strongly implied I was guilty of plagiarism or copyright infringement.  When I contacted them to see what the problem was, all they did was send me a list of their rules and tell me I needed to comply.  Since I could not see any rule I had broken and since they could not tell me what the problem was, there was no way I could comply.

I removed all of my pages from their service and went elsewhere.  That company is long gone now.  And good riddance.

I did use public domain content on that page.  They had no rules prohibiting using public domain content.  I didn't claim the public domain content as my own, so plagiarism doesn't apply.  Since it was public domain content, copyright infringement does not apply.  I don't know if either of those were an issue since they would not tell me.  They did not ask for proof or documentation that I had the right to use any content; if they had, I could have shown them where it was public domain content.  All I can do is guess and there is no way to bring something in compliance with rules when you cannot find out what rule you allegedly violated.

I mean, these huge tech companies have bots and AI and whatnot to write messages and even try to write novels and stuff, yet we are expected to believe they are completely unable to identify specific errors or specific rules that have been broken?

Despite whatever they say, you are not a partner or valued team member or part of their community or whatever flowery prose they use to make you feel all special and part of something.  You are nothing to them and, given time and opportunity, they will treat you like that.

IMHO, that is another reason why people need to not rely on or build a livelihood or anything around any specific tech company or companies.  Always have alternatives.  Do not be reliant upon Amazon or Facebook or whoever to reach your audience/customers.  They will always have the upper hand and they will use it.