That's not quite true because of private functions, which are only checked in the file where they're defined, but yes, there are improvements to be made here.

@belkadan The part about redundant checks when we have a number of overloads is true, right? It seems I badly formulated the issue.
I am going to try to change the logic to detect redeclarations the first time such a pair is compared. I've also encountered something strange: when I test with two functions, they are both present in the DeclMap from the first check. With two or more variables, they appear gradually, one by one along the way as if nothing after the current decl is visible. Is this the way the type checker works? This is the reason the issue doesn't apply to variables.

OK, now I see this is a double-edged sword. Some variables may yet not even be part of the AST when 'checkRedeclaration' is invoked, functions are often not type-checked and thus don't have interface types, especially in combinations when the current decl comes earlier than others (a, b). Since we cannot guarantee that everything is type-checked, there is little place for improvement and the function will have to be invoked several times anyways (Everything might finish type-checking only when the last decl in this context is checked by this function).

Still, I can change this to catch redeclarations at (a, b) instead of (b, a), but I am not even sure there will be a use case. Everything that I've tried ends up in 'b' either being not type-checked or not present in lookup results.