First, it has nothing to do specifically with the mechanics of the HTTP protocol. There was an emergent pattern as people built APIs over HTTP, that they can re-use much of the semantics in common with HTML. HTML over HTTP naturally led to the definition of REST, where forms, semantics, and hyperlinks are not just an optional feature but a necessity.
Formal specifications, meaning hypermedia API media types are just emerging, Hydra[1] and Micro API[2] to give examples. Fielding wrote about REST in 2000 and about a decade later was rediscovered by industry, and I hope that two decades later, people will rediscover its utility for APIs. My interpretation is that Fielding did not "invent" REST, but rather formally described emergent behavior on the web at large. Implementations may have differed wildly but had many features in common.
The problems described in the article are the result of not following a media type suited for APIs. HTML is a wildly successful media type for machine-human interaction, there's no reason why there can't also be one or a few for machine-machine interaction.