The status column answers where the application is now, and the separate decision record is what can answer how it got there. One point that is easy to understate is that ActorId and the policy version have to come from trusted context — the identity already established, and the rules that were actually applied — rather than from a field the caller can set, or from whichever policy is current at investigation time. If those two values can be rewritten later, the decision table can look complete and still fail as a trail. Keeping the previous state, the new state, and the reason on the same record as the change also makes a later review less dependent on reconstructing the sequence from the status column alone.