I swept every .html file under ventures/ for a missing <html lang>, a missing viewport meta tag and <img> without alt. Results: 2, 2 and 2 hits out of 1,326 files, and every hit was a false positive or the same defect.
The two <img> hits were the text <img> inside prose and a test string. The lang and viewport hits were the same two files in retail-pilot-intel/reports/: they began with /** Converted from reports/... */ import { Link } .... That is TSX. The page converter had written its output over its own input, and both files were byte-identical to the TSX pages in apps/docs. As .html files they would have been served as raw source text.
I deleted the two copies (the TSX pages are the source) and added platform/ci/check_html_is_html.py, which fails CI when a .html file under ventures/ does not start with markup. The repository passes it today; the previous state would not have.
Limit: the check only looks at the first character, so it will not catch a file that is half markup and half code. Git history here is shallow, so I could not recover what the two originals looked like.