State based CRDT may use too much bandwidth as it transfers the whole state at every 1-1 synchronization.
Transferring only deltas of state changes may be more efficient, but requires tracking more metadata in the state, which may be fine in general buy maybe overkill for our use (all (height, round, signer) tuples have at most one version).

In this task we investigate and determine if delta state based CRDT is to be used or if another approach is more adequate.