This argument is pretty weak when there is no real limit on the ability of authors to just keep churning out posts and getting rewards from those new posts. The computational burden of one post that can earn for a long time vs. a series of posts one after another is mostly the same, and even favors the former.

It is more a matter of how the code is structured than inherent computational costs. And indeed that does matter in that there aren't infinite development resources and many improvements that are strongly desired (communities, SMTs, etc.)