Investigate sticky sessions to allow per-pod query caching

The following discussion from !700 (closed) should be addressed:

From slack

Bohdan  [7:06 PM] answering how it currently works. all our load balancing is random, both on ILB to pick the node and on kube to pick the pod. if we need stickiness we need to do it at logic level using reverse proxy to check headers

Michael Usachenko  [7:07 PM] is that possible?

Angelo Rivera  [7:10 PM] I feel like there are subtle gotchas with stickiness

Bohdan  [7:10 PM] yes, but would need time to explore: rails needs to pass session id or cookie to the grpc call need to figure out a proxy which we can use and how to configure [7:11 PM]we can well deploy our own redis in such case