Follow closely CRDT

More info here: https://serverless.com/blog/crdt-explained-supercharge-serverless-at-edge/

DBJS data replication across instances is already implemented mostly according to CRDT rules.

Still it doesn't support operations as increment and insert (technically in dbjs collections are sets, so are unordered, although order is guaranteed and follows operation timestamp. It could more adequate to apply depth first traversal as defined in CRDT).

Additionally it's not as intelligent in handling unordered set operations on collections, where we replace whole collection (in below case a map) with a new value. e.g. having following order of operations

1. Client A issues operation A1:foo = { mar: 1, bar: 2 }
2. Client B receives operation A1
3. Client B issues operation B1: foo = { mar: 4, bar: 7 }
4. Client A issues operation A2: foo.mar = 3
5. Client B receives operation A2
6. Client A receives operation B1

Will result in foo being { mar:3, bar: 7 } in both clients, when it should be { mar: 4, bar: 7 } (as foo.mar = 3 edit was made on stale version of object, which in a meantime was replaced in context of B)