My Accessibility Scanner project started when I was still working at GitHub and I wanted a better way to catch Web Accessibility issues. The current method seems to be to lint for issues in HTML but in a world where the final HTML that is delivered to customers is built over thousands of partials and components it makes it hard to make sure that every rendered version of the page is accessible. My solution to this problem is to scan for accessibility issue in production, that's where users are hitting these issues. We ship the scanner to production and scan the page as the user navigates. When I tried this initially with a existing scanner it wasn't feasible for performance reasons so I set out to write my own scanner, how hard could it be? I don't work at GitHub anymore so development has been put on the back burner in recent years but I've been working on implementing it at $CURRENT_JOB. It's almost ready to ship to users but it needs some more testing and fixing.
I really believe that this way of treating Accessibility violations as production errors is the way to go to truly discover accessibility issues in your web app.