Maybe I'm missing something, but where does fail2ban come in on the "keys vs. passwords" debate? Do I become more vulnerable if I run fail2ban on a keys-only ssh setup? If not, then why is the argument "don't use passwords or fail2ban" rather than just "fail2ban doesn't provide adequate protection for allowing passwords on ssh?"
Same way you manage any other keys - configuration management software. "Keys are hard to manage" is not a valid argument against "you should not be using passwords or fail2ban".