Tracing a Raft Leader Election

A 5-node Raft cluster (S1–S5) has S1 as leader in term 4. A network partition splits the cluster into {S1, S2} and {S3, S4, S5}. All nodes have identical logs up to this point.

1. What happens in each partition over the next few election timeouts?
2. Can S1 continue accepting and committing client writes during the partition? Justify with the quorum rule.
3. The partition heals. S1 is still sending heartbeats claiming leadership in term 4. What happens, and why does Raft's term mechanism make this safe?
4. Suppose instead the cluster had 4 nodes total, split 2/2. What is different about liveness?