When a candidate receives an AppendEntries RPC from a server whose term equals the
candidate's current term, but the request causes a log conflict, the candidate rejects
the RPC and remains in candidate state instead of transitioning to follower.
This can lead to the candidate eventually timing out and starting a new election at
term+1, which forces the current leader — who is operating correctly — to step down
unnecessarily.
Observed in: cluster membership change scenarios where a node's log is behind the
leader's (e.g., newly joining or recovering nodes).
Affected component: candidate AppendEntries handler