This approach solves the problem at the Kubernetes-level. That is, Instant Connect (and other Kubernetes applications) are unaware that they are being distributed between different locations.

Nodes in a cluster expect to be able to communicate with one another using a high-speed, low-latency network link. Similarly, persistent volumes (disks) will need to be replicated across this link, too. Achieving this level of network connectivity between physical sites may be difficult or impossible.

For the same quorum-related reason that a cluster cannot have only two nodes, a single georedundant cluster must be distributed across three (or more) data centers. Some solutions achieve this by running two primary data centers and hosting a third, tie-breaking element in a small 'closet.'