What I'm saying is that this:

  if (typeof obj === 'string') {
    // obj is a string
  }

is really not anything meaningfully different than this

  if (obj is string as obj) { // as in C#
    // obj is a string (shadowing the outer scoped obj)
  }

or this

  if (obj instanceof String obj) {
    // obj is a String here
  }

you can imagine a pseudo TypeScript language like this

  if (typeof obj === 'string' as obj) {
    // obj is a string
  }

Actual TypeScript merely allows you to elide the "as obj" and all it needs to trigger this is the "typeof obj === <string literal>" inside of an if expression. This can be done by a simple syntactic replacement, it doesn't require control flow analysis to get this specific feature. But yes, if you have a more general computed expression, that would apply, but that was not the case you were stating. Ie the variable isn't the issue (redefining in this case seems like a distinction without a difference)... However this does work in TS, which is a demonstration of CFA:

  const isStr = typeof obj === "string";

  if (isStr) {
    // obj is str
  }

For what it's worth I think CFA is useful in TypeScript based on it at its core being a structural typed bandaid over JS, but I think these specific CFA type narrowing features are redundant in stronger typed languages.

Java cannot do the same without breaking backwards compatibility, hence the idea to introduce a variable explicitly.