I was under my own (pretty pedestrian) assumption that the secret sauce of JB IDES is/was the MPS system in which they are/were able to create libraries for source code analysis and manipulation on a language-by-language basis. I further assumed that this per-language library plus some language-specific GUI tooling, plus the language-independent GUI libraries and processing (local history, VCS integration, side-by-side textual (not semantic) diffing tools, and so on)

The response I got, which was quite a while ago (2019 or before) was that the individual language IDEs did not use MPS-generated libraries in perhaps the extensive way I've alluded to above, and this surprised and disillusioned me somewhat. I was hoping to get as much of an answer from my JB friend as, presumably, you were from me in this thread. :-)

Yes of course, there must be massive code reuse between projects, and yes, this was an entirely hand-waving set of assumptions, and no, I haven't researched the composition of any given JB IDE to find out the proportions of language-independent vs language-specific code volume.

I'm surprised someone at JB hasn't given a talk on this sort of thing over the years.