Nope; latency would be about the same, since it requires two rounds for consensus for all ops. Note that Raft itself has no clue about the specifics of the op, whether it is a read or a write, and for leader to be sure it is still the leader, it needs to wait for an answer from a majority of servers. Of course, this is not done for every single op, but batched. Similar methods will have to be used for my suggestion as well.

Now, in many production cases, one tends to put a lease on leader election. Until the lease is valid, the leader is guaranteed that it is still the leader and no one else has a more up to date state. This allows reads to be served from the leader's state machine without having to touch Raft at all.