It's also horrific because you can use that to brute-force guess someone's password, because the robot will tell you when you have the wrong digit, so you can work your way through all 10 phone keys for each digit, noting each time whether the robot kicks you out, until you guess the entire thing.

To be clear, the input just times out, likely because they don't expect a long random password input. I don't think it's evidence that they can verify a substring of your password.