Design a theoretical consensus algorithm for a distributed network of 100 nodes where network partitions are frequent but message duplication is impossible. The system prioritizes availability over strict consistency (AP in CAP theorem). Describe the leader election process, how log replication is handled during a partition, and the mechanism for resolving conflicts when the partition heals. Compare your design's efficiency to Raft or Paxos in this specific context.