How does one handle hotspotting of certain keys in a redis (or other in-memory cache) cluster?

Question

By "hotspotting", I mean certain few keys ("hot keys") being accessed far greater than other keys.

If keys are distributed across a cluster of nodes,