leader-election - Question and answers

I'm building a distributed system for a real-time analytics application where nodes need to elect a leader for coordination. I'm using the Raft consensus algorithm, but during peak traffic, I've observed split-brain scenarios where more than one node claims leadership, causing data inconsistencies. I've set timeouts and retry mechanisms based on the Raft paper, but logs show network delays might be a factor. What steps can I take to stabilize the election process?