So if there is some superior notation that would benefit all of maths, that sounds very interesting but I don't think you've made that point very well. Maybe a better example is needed, because on its own, eliding repetitive information from notation does not seem like a problem.
With how the notation is typically used, it’s at least closer to my mental model.
Treating the expression as a function under substitution is one way to look at it, but not the only way.
Also you don't need a symbol type to deal with variables, you just need to allow type variables to be used in standard expressions. (define x R) (^ x 2) is just as easy to manipulate as (λ (x) (^ x 2)), the only issue is that your rewrite rules how aren't just looking at syntax, but semantics too.
I don't mean this in the usual "we need to invent new symbols to make it clearer" way.
I mean it in the way that it has implicit typing that gets coerced constantly. Using Haskell types D :: (R -> R) -> (R -> R). Yet it gets used on things like D 2 = 0 which implies D :: (R -> R).
What you've actually done is an implicit conversion of 2 :: R to 2(x) = 2 :: R -> R and 0 is not 0::R it is 0(x) = 0 :: R -> R.
The Q_0 system described in [0] is a good start for a mathematical system that is both sane and as expressive as the one we have now. With term rewriting you can even keep the ridiculous notation we have today if you love it so much. Though I have grown incredibly fond of typing out extended typed lambda calculus using scheme notation in Emacs with a home grown automated theorem proving mode that was trivial to implement.
[0] https://www.springer.com/gp/book/9781402007637