First, television and movie production is not "agile". It's waterfall. The script is the design.
Live theater is "agile"; theatrical directors go into a rehearsal room with a group of actors to have a debug session. What comes out may be very different from what went in. In that environment, "I know it when I see it" can sometimes be found by trying different things. You do that in TV or movies and the cost goes through the roof as the production people cope with the changes. Theatrical directors who move to Hollywood sometimes have real trouble with this.
This used to be less true. As projects became more set and effects heavy, more and more is locked in before the first shot. Hence the "previz" business of making the project twice, once in a crude storyboard form and again for real.
Second, there's no patching the product after release. There's no "pivoting". You only get one chance to make a first impression. Although sometimes, movies and TV shows are tested on test audiences and then re-cut.
What the article rails against the most is blind iteration, the "make it fun," or "make it pop," "just change this one thing," type of vague sentiment that critiques but doesn't really direct. Blind iteration happens because the leads(or even on a solo project, the author as lead) ran out of templates to follow, and when you run out of templates you have to begin using trial and error to get results. But it's immensely cheaper(to the point of it being required in many types of project) to focus trial and error around finding a new template with better, clearer rules, than it is to send the order to make something to finished quality and then decide that it won't work and start over. However, the latter tends to happen every day, in all sorts of scenarios, from a poorly considered dinner plan to high-profile software projects.
Anyway, who says a startup has to be "agile"?