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....
