The laundry list of problems that can land in one's lap going down the road of XML publishing is considerable. Many are non-negotiable, particularly on git-based version control or on modern computing environments with locked down Java versions. It would be boorish of me to cite all of them, one after the other, so try this exercise instead:

Using an XML specification of your choice, working from a simple desktop computer, and with no money for fifty thousand dollar "end-to-end publishing solutions", perform the following: publish a document that uses conditional processing to show two different outputs for two different audiences while using a shared set of document components.

<kitchen_timer/>

What did you end up using? Anything neat?

I'm going to bet you don't find anything at all.

The tooling in XML publishing is weird, locked down, horrible, or all three plus some. Plus a lot. So, even if we pretend XML publishing is the bee's knees, it loses the speed test, it can't work on vanilla computers/requires admin rights constantly, and new writers train up on Asciidoc, from zero, in four hours. At a glacial, defense-industry-standard-for-fifty-year-old pace. Git probably provokes more existential crises in training than the markup does. Nothing compared to the business side, of course, but this is the defense business, and it's all madness.