Redis slaves cache rate mismatch

We have set up 3 Redis nodes amongst which 1 is designated master(handles writes) other 2 are slave nodes and act as read replicas.We have 3 sentinels also configured for automatic failover.The client apps are all Java based applications using Lettuce library.

We are observing a huge variation with the cache hit rate of both the replicas. One of the replica has a cache hit rate around 96% while the other one is around 77%.At any given point both these nodes have the same number of keys.

Any reasons for this and how can we debug this behaviour?