As far as your first question is concerned, macros in and of themselves are not incompatible with a typed language. Coalton uses Lisp macros, for example, and they work seamlessly and as expected.

But a Coalton macro is typically written in Lisp, not Coalton. This isn't in and of itself a problem---it's very easy and straightforward thus to write Common Lisp-style macros in Coalton. The difficulties arise when the macro must itself be (1) written in a statically typed manner, which gets into having a complete typing on your metalanguage (or your AST), and (2) allowed to access the type environment of the surrounding context in which the expansion is happening. If (2) should be accomplished, then the type-checking machinery must collaborate with the macro-expansion machinery, and that in practice makes both tasks very difficult to specify semantics for and implement.

The language Hackett [1] worked toward solving the problem of having true typed and type-aware macros (they call them "type-aware" and "type-directed" macros [2]). Development unfortunately ceased ~7 years ago.

[1] https://github.com/lexi-lambda/hackett

[2] I think this video has a good discussion of Lisp, macros, and static types, from the perspective of implementing a Haskell in Racket. https://www.youtube.com/watch?v=5QQdI3P7MdY