I really much like the idea that free software allows to keep or even revive old version, and really dislike the modern trend of “keeping *updated*”, with the exceptional care about security (well-done and simple programs like TeX can avoid update and stay secure and bug-less by being simple and good enough: lacking update necessity is a really good feature).
Currently indeed anybody would argue old versions are bad and newer better, because of course newer is what *current* main developers use and *like* (maybe it allows easier development on *some* aspects *they like*). But the tons of bloated new features you mention seems to me not that universal, so superiority of newer over old seems less obvious. So, if you dare, please do and revive the old versions, update them, make new minor releases, and possibly keep that compatible with old API and programs… possibly even adding new cool (and not bloated or common with newer major releases) features, without breaking backward compatibility. More specifically, for instance, it is sad the MATE desktop switched to GTK3 (afaik). Maybe (afaiu) they did because of security, build issues, or other issues you could solve, so they could go back back. Even if not fully completed or appreciated, consider such work will anyway teach you (and possibly others, studying your work) how these (and more generally: such) projects do work. I recall the statement of someone (I think it was José) you said one of the first versions of GCC is really well written, small, self-contained and thus easier to understand than current one, so it’s very useful to quickly understand current state of its internal architecture by giving a grasp at it… that might as well apply for this.
