Optimistic reconciliation allows, multiple update of shared data without synchronization. The assumption is that the vast majority of the actions will not conflict. In these systems, write availability is raised in the presence of network failures, high latencies or parallel development. However, in order to remain consistent, optimistic systems repair divergences. To produce a new consistent state, they use the logs of each user in a process called log-based reconciliation. The purpose of an efficient reconciliation engine is then to compute a new consistent state which preserves the maximum of previous actions. This work leverages the efficiency of constraint-based reconciliation. It provides a new efficient two step procedure built by connecting theoretical results on backtrack-free search with this hard optimisation problem.