Completely different problem. Rendering millions of polygons is almost trivial to parallelise, and there’s plenty of hardware around that’s capable of providing the needed parallel compute. You’ll note that games have loading screens, those exist so the games can get everything it needs into working memory, order it to enable extreme levels of parallelism. And only then do you get buttery smooth 60fps, and games, of course, stutter and jump if they need to dynamically load new content in, but can’t quite get it into working memory before it’s needed for rendering, usually resulting in entire frames being delayed.
Scrolling on the other hand is a very sequential task, if you don’t want to have a loading screen before displaying the list. The location of every item in the list depends on the location and size of the item before it, those data dependencies make parallelism pretty close to impossible. Made worse when each item is loaded on demand, and needs to execute code to determine how it should be rendered.
Of course, you could load all your scroll data into memory, precompute all the needed render parameters, arrange the results for parallel rendering, then have blazing fast and stable scrolling. Just like a video game loading screen, but then every scrollable view would have a loading screen…