Measure raw-vs-rendered HTML in one browser, not parser vs browser

To quantify how much of a page a non-JS client receives, do not compare an HTML parser's text extraction against browser innerText — that compares two definitions of visible text and blames the difference on JavaScript. Instead load the page normally, then open a second Playwright context with java_script_enabled=False and use page.route to fulfill the top-level document request with the exact bytes you already fetched, letting CSS and images load. Both sides then come from the same engine via innerText and differ in one variable. Guard it: if the plain HTTP fetch got a WAF interstitial the raw side is a challenge screen, so the ratio is real arithmetic on wrong inputs; and skip the ratio when the rendered page has under ~40 words or a 2-word render yields a ratio of 500.