Many years ago smart frameworks implemented smart stuff like you can display only what is visible. For example you could have a table with 1 million rows but in your html page you will not create 1 million row elements, you can create GUI widgets only for the visible part, as the user scrolls you can recycle existing widgets.
As a practical example , you go to a yotube channel page and they load only 2 or 3 rows of videos and you have to scroll to force more to appear, this means you can't do a Ctrl+F and seatrch and is also less efficient because as you scroll the items at the top are not recycled and reuse so probably more memory is used.
The json for all the videos is not huge,some strings with title and thumbnails, maybe some numbers but the issue is that is not possible to natively do the best/correct thing, only recently we got lazy loading for example so basicaly html was desibned for documents and frameworks/toolkits designed for apps did the correct thing many years ago... this is an explanation but no excuse why things are such a shit show with pagination today.
For the main content, yes, yes, most will. Why are they on the website in the first place?
Especially the main content. Sure, some things you can load later, like comments, etc