But I'm not making an argument am I? I'm making a question.
Whether you are aware of it or not, you are making a number of assumptions, some of which come off as argumentative and/or coming from a potentially biased standpoint. That said, the most useful thing is to look at what has actually happened with various similar games.
Some instructive case examples to be looked at are KSP (Kerbal Space Program) and Stellaris.
KSP is similar in that it was a very small division of a small company, that self-published for a number of years while describing their game as alpha / beta / early access, while in practice a large number of people found it already enjoyable and were treating it more as a released game. As a highly customizable game that was on the borderline of games and simulators, they had mods focused on small tweaks, major rebuilds, user interface improvements allowing more specific control of the world's elements, and all sorts of graphical enhancements for those who thought the original devs were not focused enough on eye candy. They eventually launched on Steam, and various other online stores. (Side note: the console versions really are different games, and not discussed here.)
Originally, the main modding scene was focused on an external site run by users, with discussion split between the developer's own forums and the game's reddit (most important material appeared in both places). That site went defunct. A replacement site sprung up in short order, also user-run and external. At some point, the developer announced that their official mod site would be a commercial external gaming site. While some mods moved or copied there, it was regarded with general suspicion.
As time has gone on, more and more mods have been managed as github projects; some release only there, others copy to one or both other hosts. The developer exerts negligible control over mod publishing, and minimal control over mod discussion on their forums; for the most part this works out fine.
A user attempt at a mod loading / management framework has been around for a while, but many mod devs don't like it, as it has a reputation for dramatically increasing support headaches due to version issues; on the other hand, a significant number of very non-technical users seem to depend on it. (The number of people who play sim/games who are not able to do things like handle extracting zip files and putting things into a specific directory is always a surprise; the average user age is less than you think and their basic computer education isn't as good as you hope.)
All in all, KSP seems to be a fairly comparable example where letting publication decisions remain entirely up to the mod authors has not resulted in any serious problems. Mods can be, and are, published on any combination of the fan-run site, the commercial site, version control hosting sites like github, and individual user websites. As mods from any source are literally the same file, it doesn't matter how one gets them. Users of the fan-created mod updater don't even see the difference (when it works), as the mod download source is abstracted away (ideally). That said, the ease of maintenance and collaboration seems to be pushing more mods to have github as their prime repository, as other sites are mere file repository / downloaders and don't have the rich versioning, developer, collaboration, and in some cases hand-over tools.
Stellaris is less similar to DF, but has some interesting comparisons in terms of mod ecosystems. It's interesting in that while Steam Workshop seems to be by far the most popular way of getting your mod seen and distributed, it's not actually required; mods from the workshop can be manually copied and applied without using it, and non-workshop mods can be manually installed. That said, few people seem to do either (with the exception of local personal mods), and AFAIK no one has put in the considerable work to develop a third-party fan-run mod hosting solution that would compete.
Moreover, the absolute firehose of mods,
most of which are utter bunk, serve to clearly disprove any ideas that the Steam Workshop is inherently a "walled garden" in any way (a busy yet sketchy interstate truckstop might be a better comparison). Publishing a mod is, arguably, too easy; a significant fraction of them posted are multiple versions of some random user's personal-use adjustment where people don't even understand that they are publishing to the world, lame jokes or short-lived memes.
(An interesting practical lesson is that a quite significant number of the mods are translations, not only of the base game, but of other mods. You even get mods that are translations of compatibility mods written to allow two other mods to more or less work together, where all four authors (mod A, mod B, mod A+B compatibility patch, the compatibility patch's translation) are working completely independently. The volume taken up by meta-mods and translations exceed that of the original mods by a noticeable fraction; some of this is due to the nature of connections between things scaling with the square of the number of things (N*(N-1)/2 generally.))
Of course, in the end DF will be its own thing. We all hope that the community around DF will be better, or at least no worse; it will undoubtedly be *bigger*, and that is a mixed blessing.
My main worries are what happens if it actually sells well at first, but gets some sort of backlash. There are a lot more companies that have gotten that sort of thing wrong than right, and skilled community managers that can actually be a net positive are rarer than anyone would like. The DF forums are not so much a walled garden, but
a very obscure park that remains a lot more peaceful and organized than much of the world that surrounds it