A few thoughts that occur to me reading that in no special order -
I'm not a believer that decisions can be made mechanically through data collection. I do believe that the right kind of data is an important input into judgment and experimentation.
I think it's wrong to say there's no data now. If a developer is reading feedback, reading bugs, looking at how everyone around them configures their desktop, running the occasional user study, talking to customers if they're at a company, using the software themselves, etc. then they are getting a lot of data. It isn't a "study" but there's still a lot of basis to identify and figure out problems.
When I was actively maintaining metacity then I had FAR more information about how it could work and why and what kind of people would like solution A vs. B, than I currently have about gnome-shell or other WMs. If I were suddenly maintaining some other WM today, I'd have to spend months immersed and learning before changing anything much.
The point is that as a maintainer, I felt I had lots of decent information to base decisions on, and as a not-maintainer, I don't feel I have that information anymore.
There's this fantasy we all sometimes have that it can be a science and/or the answers are clear and cut-and-dried and/or that the developers could "prove" to us that they are right. But it's just too complex for that.
The alternative to "completely mechanical" is not "completely random/subjective" though. There's still a lot of room for good or bad judgment, based on good data or not. "Judgment" is not the same as "subjective opinion."
More data can always be good (with privacy protections in place, wouldn't it be great to auto-collect from all Fedora users which config options they had changed and what desktop they used, for example? that would be handy and 10000x more valid than a web poll).
Again: I'm not trying to get into whether project XYZ's judgment about topic ABC is _correct_. Judgment can always be wrong. What I don't agree with is what many imply, that judgment is the wrong _process_ and could be replaced with something more mechanical.
On the specific issue of whether "the silent majority wants this" can be an all-purpose excuse: it certainly could be used that way. When I was daily immersed in metacity-related data, I think I had a reasonable basis to know what was of general 80%-ish interest and what was one or two people or what was 15%-ish of people. So I would think the current GNOME maintainers also have some decent data on that front. But _could_ they ignore the data, use "silent majority" as an excuse and not really think it through? Of course they could. Do they? I don't assume that by default.
Perhaps the make-or-break marketing thing for Linux desktops is to be included in distributions, especially by default. If GNOME were a for-profit company and I were CEO, then getting Ubuntu back on board would be on the top of my list. Distributions are where the ... distribution ... is.
A little tangent ... lots of times "confusing" is not a good way to think about UI-goodness. I say it too, but "relevant" or "appropriate" would be better metrics.
One reason is that "not confusing" does not mean "good" (problem = "I really quickly understood how this works, but I don't want to do what it does"). "confusing" does not always mean "bad" either ("oh, now that I get it, this will save me hours of time"). The "confusing" dimension gives little direction. (This is a limitation of "how long to do the task" user studies...)
Another reason is that "IQ" or even "computer savvy" isn't the relevant distinction among the groups. It's more like "what previous UIs do I have experience with" and "how much do I like to mess with the computer vs. focus on other things." I really dislike narratives around "power user"/"dumb down" because they imply some type of progression or ranking, when what you really have are people who care about different priorities or have a different set of life experiences. To get a relevant/appropriate design you have to think about that.
At one point long ago we were trying to make the "tweak tool" UI be a "people who are used to Unix" UI, like "Unix options" or something. That had some other pros and cons I don't remember, but one "pro" is that it makes clear what the UI is supposed to achieve, while "it's for power users"/"it's for confusing stuff" really does not. If it's framed positively like that then you can start saying "what are all the things Unix users will be used to," etc. and try to get that stuff presented nicely.