A defensive programming nit is not to return the desired result [1] but to show all collected cases are true.

  def defensive_true():
      # assume False
      ...
      return arr.all(True)

There are tests, but some wild refactoring in the night and "tests still pass," leading to a blind ship, could be a tiny risk.

Not a big deal, just I don't trust myself to return the happy path if the middle could change.

[1] https://github.com/teorth/estimates/blob/main/src/estimates....