yes. 1) it would be significantly less secure - compromising either device would compromise both. Imagine an airplane with two engines that needs both to fly - a single engine plane is actually safer - because the chance of loosing one of one is less than the chance of loosing one of two, (assuming chance of engine failure is independent) Use a separate key on each device is like a two engine plane that can still fly with one engine - this is significantly safer than a single engine plane.

2) it would greatly complicate the replication protocol, having to take into account forks, rather than assuming append only, where you can represent the current synced state with a single counter.