Rendering strategy for a large content site

A content site has 2 million articles, 40 million monthly readers, and a personalised "recommended for you" rail plus a signed-in header on every page. Editors expect a correction to be live within 60 seconds. Design the rendering and caching strategy. Cover what is static, what is server-rendered, what is client-rendered, how the personalised parts avoid poisoning a shared cache, how the 60-second correction propagates, and what happens when the origin is down.