In Raft, the leader queues and processes requests in order rather than refusing them. Techniques like batching, pipelining, and asynchronous handling allow the leader to maintain high throughput and low latency, ensuring clients get a timely and consistent response.

Example:

Let’s say a leader is processing request1 when request2 arrives:

• request1 is added to the log and sent to followers.
• While waiting for replication acknowledgments, request2 is added to the log and queued for replication.
• Once request1 is committed, the leader begins applying it to the state machine and simultaneously replicates request2.
• request1 is applied to the state machine, and the response is sent to the client.
• The process repeats for request2.

You can see its implementation here https://github.com/dhyanio/discache/blob/main/rafter/raft.go