I’m not a fan of motion-action. Most of the time in vi, all I’m doing is moving around. When you input a motion key in vi it moves you accordingly with no fanfare.
The motion-action paradigm, on the other hand, primes the editor to expect an action. In many cases this leads to multiple cursors being left in your wake, which I find really distracting.
I think this stems from a fundamental misunderstanding of the vi paradigm: the primary purpose of vi was to extend the traditional line-editor ed (which was designed for editing through a paper TTY terminal) to full-screen interactive editing on a video terminal. Since early video terminals had very low baud rates, vi needed to be very efficient at reducing screen redraw. It just so happens that this emphasis is perfectly matched to the most common editing loop: jump to a location within the file and insert/delete/replace some characters. Thus vi happens to be perfectly optimized for this editing workflow and so I believe this is the reason for its enduring popularity.
The other task which vi is strong at (but which is otherwise secondary) is editing in the large: making multiple changes to a file simultaneously. This, I believe, is the primary strength of motion-action paradigm editors. In other words, they get it backwards! Thus they can’t help but annoy me while I’m working with their constant attempts to infer large-scale edits from my simple desire to jump around the file and make small changes. They are giving me a freight train when all I want is a car.