While the documentation explicitly warns about the "semantic disconnect" for non-visual users (screen readers), it may not fully address the "visual disconnect" for low-vision users who rely on magnification.
Examples might include
- Emphasis on keeping elements in viewport - not clear if highly zoomed in users will see an anchored element (like a tooltip)
- Focus tracking seems visual - if anchored elements appear without focus, there won't be automatic panning to the new content