Personally I prefer infinite scroll, versus the alternative of finding the "next page" button at the bottom, waiting for the content to load (preloading could help here) and sometimes navigating to the beginning of actual beginning of the content I was viewing. I even used a browser extension that matched "next" buttons from pages and loaded the next page content automatically, but the extension (can't recall its name) is not available anymore.

Granted there are some downsides, such as having the browser keep extra-long pages in its memory, but overall I prefer working infinite scroll mechanisms over paged ones. As far as I see, the ability to remember the current location in the page could be easily implemented by modifying page anchor and parameters accordingly, though personally I've rarely needed it.

Perhaps if there was a standard way (so in the HTML spec) to implement infinite scrolling, it would work correctly in all cases and possibly even allow user to select a paged variant according to their preference.

Not all the paged views work correctly either. In particular systems that show threaded discussions can behave strangely when you select the next page. Worst offender is Slashdot.

Infinite scrolling, at least in most implementations, makes it almost impossible to find historic content. This could be useful e.g. to journalists and researchers, but even as a private individual I would sometimes love to be able to see what person XY posted in, say, 2019.

In other words, infinite scrolling works in "pure consumption" contexts. You'd never want to add infinite scrolling to some backoffice admin interface.

1 batch of content = 1 batch of add space = more money.

Each next page click is a moment for you to reflect and notice the waste of time. Simple as that.