For full-page captures, I'm using Chrome's native DevTools screenshot API (the same one you get with Cmd+Shift+P → "Capture full size screenshot"). So the behavior for sticky headers, lazy-loaded content, and dynamic pages essentially matches what DevTools does—both the benefits and limitations.
Sticky headers: Captured in their fixed position throughout the scroll, as DevTools does.
Lazy-loaded content: Depends on how Chrome's capture handles it. Generally works well for standard lazy loading, but infinite scroll or heavily JS-dependent dynamic content can be hit-or-miss. That's a Chrome limitation rather than something I can work around in the extension.
Dynamic resizing: The viewport setting works well here since it's part of the DevTools protocol. Pages render at the specified dimensions during capture.
For visible area captures (not full-page), I have more control and it's straightforward—basically a direct screenshot of what's rendered in the viewport.