Riak version used is 2.0.6 EE. {Fred} -> {Bob} -> {Jane} {Fred} has: • Bucket a: n_val = 5 • Bucket b: n_val = 3 {Bob} has: • Bucket a: n_val = 5 • Bucket b: n_val = 3 {Jane} has: • Bucket a: n_val = 5 Sequence: 1. I set {Bob} to have cascade writes = true 2. I created a new bucket on {Bob} [ABTest] and set the repl property to false 3. I then inserted a new object into [ABTest] on {Fred} 4. A get on all 3 clusters returned the new object 5. I repeated the same test, by updating an object, and deleting it It looks like cascade write causes the repl flag of a bucket to be ignored (when it is being interacted with via replication)