What is a common mistake with availability zones and regions in Cloud Architecture?

Calling an application highly available while all critical dependencies remain in one zone creates hidden single points of failure.