Let's call "epoch" to each period with a given configuration of valid private shares (each of them held by a different node). On each successful recovery or refresh protocols, there will be a new set of valid private shares, and therefore, a new epoch.

For the moment, we don't have recovery and refresh protocols in production, so all our cohorts can only be at epoch 0, but for next releases, we will need some versioning mechanism to handle new epochs.