Websites that use reasonable or good HTML markup with structure, the correct HTML tags, useful ids and classes are great to work with. But regularly you see websites with generated HTML without any useful identifiers or structure. A generated garbled mess of anonymous, generic components and styling CSS classes.

I've worked on content extraction for OpenTermsArchive and write my own injected CSS hacks and browser extensions. Working with good website sources is great. Working with garbled messes is awful.

HTML losing its markup aspect - that you can traverse and select - makes websites inaccessible.

/edit - adding:

The CSS tailwind generates might not be bloated, but repeating the gigantic strings of classes all over your codebase certainly adds to the size of the final HTML output.

The HTML is not just bigger, but bloated and inaccessible. HTML markup with identifiers and classes is readable and understandable. It has structure and labeling. Inlining styling rules bloats it to the point of unreadability. And losing identifiers and classes is a loss of labeling and selectors.