Why server-rendered pages look ready before they are. Measure hydration cost, avoid mismatches, and reduce it with islands, progressive, and selective hydration.
Server rendering gets HTML on screen fast. But that HTML is inert: buttons don't respond, menus don't open, inputs may lose what users type. It becomes interactive only after hydration:
Between "HTML painted" and "hydration done" the page looks ready but isn't. Users tap a button and nothing happens. This gap is the uncanny valley, and it shows up in metrics:
The key insight: hydration cost scales with the amount of component code on the page, not with the amount of interactivity. A 3,000-word article with one "like" button still hydrates every paragraph component, in a classic SSR setup.