I don't think inaccessible products are accidents. They're decisions, silent or otherwise. Usually though, they're decisions nobody labeled as such.
"We'll fix it later" has an expiration date. Because most accessibility problems aren't technical, but an issue of prioritisation, the more you postpone the more it'll hurt later. The people most affected are least likely to be in the room when things are prioritised. And that's the whole problem.
That's how you create accessibility debt. And it's real debt. And it compounds. But unlike financial debt, nobody's sending you monthly statements.
You can shout from the rooftops that you care deeply about your users while still shipping something a keyboard can't work with. You have good intentions, sure, but good intentions don't matter.
The fix is usually embarrassingly simple.
It's the stuff that takes minutes if you do it while building and hours if you go back for it later. The "fix" isn't some big dramatic refactor. It's just doing the small thing at the right time. Which means the problem isn't skill or resources, but habit and awareness.