> there have been 17 commits pushed to master by 7 authors[1].
And there are 17 thousand commits overall, yet a lot of fundamental issues remain unresolved, so I guess I'd argue that in many cases a program can NOT be substantially improved without substantial changes (again, not universally!, some crash/perf wins are "free" in that they can change nothing in the program's user-facing behavior)
So I see a lot of reasons for the shell/vim to substantially change, and thus stability in this regard is a negative, not a positive, hence my comment that it's better to use tools that improve better (though you're right that improvement is not a given for a change)
And a program can be improved even without it being changed substantially.