Usually, you get exactly what you pay for, so when you skimp on device redundancy, that's exactly what happens.
Server virtualization has brought us efficient and cost effective disaster recovery options. Snapshots, v-motion, and storage replication are the underpinnings of this redundancy. However, its incorrect to apply these server techniques to virtualized Application Delivery Controllers (ADCs). The constraints of memory utilization, product licensing, and time/complexity-to-recovery are nuances which uniquely apply to ADC devices and keeps them from enjoying the same methods of redundancy that virtualized server workloads leverage.
A redundant and available architecture always starts with TWO appliances. It's just not appropriate to architect an enterprise quality high availability solution with less. It might look good on paper, but the recovery time will clearly exceed estimates- and likely won't work at all.
In some business continuity architectures, 3 devices may be appropriate. And in the future, Spotted Clustering, across multiple data centers may change best practices, but that's still a ways away and will require additional licensing. In the mean time, the best practice solution should always start with a pair of devices in each data center.