It's plausible that Amazon removes unhealthy servers from all round-robins including DNS. If all servers are unhealthy, no DNS.
Alternatively, perhaps their DNS service stopped responding to queries or even removed itself from BGP. It's possible for us mere mortals to tell which of these is the case.
Chances are there's some cyclical dependencies. These can creep up unnoticed without regular testing, which is not really possible at AWS scale unless they want to have regular planned outages to guard against that.