The only part of our deployments that are not HA and distrubuted over multiple ndoes is the single redis pod deployed per site.

Once PLANET-4577 is merged and stable, we should also deploy a HA redis service.

There are a few options, with varying degree of complexity / tradeoff.

1. HA redis per site: most similar to current deployment, most duplicated memory consumption
2. HA redis per cluster: a single shared redis instance per cluster might be possible if we segment each NRO to its own databaseID (will involve significant performance / script update / reliability testing)
3. HA redis managed service  https://cloud.google.com/memorystore/  multitenancy likely the only option as cost per deploy would be prohibitive.