I’m familiar with the history of TeX, Knuth, the cult, etc. I was a veteran software QA manager before I became a social scientist, and I don’t like to make excuses for bad software. Open source is no excuse for software of this quality (especially when open source is often touted as a factor that leads to superior software). I have a lot of experience trying to get Miktex to work as well, mostly unsuccessful.
TeX is just an awful software ecosystem, beyond bad. Half the time, like with XeTeX this morning, there’s no way to uninstall it. It doesn’t register itself properly as an application in Windows 10, even though it was installed with an installer (of its own design; not simply unzipped binaries), so it’s not listed in Programs as something that can be uninstalled. If you’re going to issue a Windows release, it’s not much effort to use the standard platform APIs for things like registering yourself properly as an application.
The 5 GB is after some shrinking, leaving out all the languages it wanted to install. The installer was too primitive to be able to determine which packages I’d need or not, much different from say the Visual Studio installer, which orchestrates a similarly enormous multi-GB install, but which gives you lots of information and a proper UI to be able to make your choices. TeX distros usually want to install packages for every esoteric field and the thesis format of Argentina’s Air Force Academy or something. It’s unclear why they do this. It’s also unclear why no one has bothered to build a normal application that consumes a TeX dialect and renders modern PDF (v1.7) files or other outputs. Surely it must be possible to do this with a 100 MiB application. (Or even a clean-sheet typesetting markup language and system, but it’s not clear if a clean-sheet language could notably improve on TeX/LaTeX, which seem like okay languages.)
I like the equation typesetting. It’s not clear why else to use TeX distros. The most noteworthy thing seems to be Knuth’s justification algorithm, but since justification has been shown to be less readable than ragged right, I wouldn’t justify unless a journal forced me to. And they tend to want Word docs anyway. I’d be interested in seeing research on how good Word’s justification algorithm is compared to TeX, and various microtypography features, kerning, etc. No one seems to have researched it, and most pro-TeX claims on those variables tend to be comparing TeX to 20 year old versions of Word, not Word 2019/2016, for instance. There’s remarkably little research except for Knauff and Nejasmic’s study of “efficiency”, which seems to have some validity problems.