You mention writing a novella as your use case, but I don't think that's a particularly difficult task for a typesetting engine. If you found typesetting a textbook with "rich" margins, annotations, and figures significantnly easier I would be more inclined to learn about alternatives.
With LaTeX I can transform my research papers to comply with a publisher's template in a few lines of code. Sure there are some annoyances such as compile compile bibtex compile, but the ecosystem is mature enough that most of these are well handled.
If you need amsmaths or APA citation, you still get those. Except now you get them without insane commands just to type Paul Erdös' name, or citing Лев Семёнович Вы́готский.
\usepackage[utf8]{inputenc}
\usepackage[T1]{fontenc}For works that have to handle multiple scripts, and for works wherein you are in control of your own typography, it's more logical and practical to work with XeLaTeX/LuaLaTeX. If you wish to use OpenType or TrueType fonts there's little debate.
If you want to go the whole nine yards regarding Unicode, there's recently been some development regarding Unicode math, with the first impetus being Microsoft and its Cambria Math to support mathematical typesetting in Word. Unicode math refers to fonts implementing mathematical symbols as Unicode glyphs, as well as some font-side logic regarding how the font renders math, afaik.
So if your publisher has a LaTeX template, that's pretty much guaranteed to work just fine when you \usepackage it in XeLaTeX context. Needless to say, except it really does need to be said, that also means full Unicode Maths support (through \setmathfont{A normal font name} without needing TeX metrics nonsense because it's 2017) for any OpenType font with maths support (Cambria Math, XITS Math, etc. etc.)
Another point is that some packages haven't been updated for Unicode, algorithms2e being one popular package that's still Latin-1, IIRC.
Then if your code will be compiled from source by someone using pdfLaTeX, your Unicode characters will cause problems.
Hence why I say that it's probably more pain than it's worth, since there are still things you need to understand to avoid incompatibility. I think a more proper analogy would be C and C++, since while XeTeX/LuaTeX is mostly LaTeX compatible, 1) you risk running into sinister incompatibilities and 2) typical XeTeX/LuaTeX code is different from, and incompatible with typical LaTeX code.
For example I can list them with: mtxrun --script font --list --all
But when I try to use "gonormal" ( \definefontfamily [myfam] [tf] [gornormal] ) I get: "The requested font 'gonormal' has no files for the 'tf' alternative, Latin Modern is used instead."
And if you need the automatic TeX style replacements (like turning `` into opening quotes - personally, I hate that, I'm writing text, not programming - then you can pass in TeX mapping instructions in [...] blocks, covered by the fontspec manual: \setmainfont[Ligatures=TeX]{Font By Name} and \newfontfamily\whateveryoucallit{Font by Name}[Ligatures=TeX])
m n
[ A B ] m
X = [ C D ] n
[ E F ] n
Achieving this in LaTeX deserves a dissertation in and of itself. \documentclass{article}
\usepackage{blkarray}
\usepackage{amsmath}
\begin{document}
\[ X = \begin{blockarray}{rrr}
m & n \\
\begin{block}{[rr]r}
A & B & m \\
C & D & n \\
E & F & n \\
\end{block}
\end{blockarray} \]
\end{document}
It's like talking about how hard it is to work with browsers and then telling people the best way to set up IE8, instead of using a modern browser.
Of course it's still you in the typesetting driving seat, it's never going to be as easy as Word, but I'm writing a novella in XeLaTeX right now, not even maths papers, and it's almost child's play to make it do what I want with minimal effort. My entire "complicated setup" is five lines plus one line that's me being completely unreasonable, with XeTeX going "sure, be that way":
And that's it, the rest of the document is basically just a giant text file of paragraph after paragraph of normal text, using normal as-per-unicode accented letters saved in UTF8. The doc's certainly interspersed with \part{} and \chapter{} markers because that's how a story works, but there is nothing I couldn't explain to my mom in there: I'm writing a "book" with a table of content, a reasonable looking page header but otherwise plain styling, with some explicit page dimensions (which is no different from using Word for a specific page size).(the only unreasonable demand being that "\input" line, which imports a set of rules for using "Palatino Linotype" as main font, but then pulling the punctuation glyphs from "Sylfaen". So stuff like full stops, apostrophes, commas, double quotes, etc. And even for something seemingly as complex as that, is just a few more lines using modern XeLaTeX)