Quick update,
last week, I completed the simplification of the election protocol, 
which is now triggered by JGroups views and it is now completely 
deterministic, while still observing the properties mandated by Raft. 
This is in 1.0.6, which I released last week.
Next, I added leader election edge tests as described in the Raft paper [1].
Francesco found a perf bug with the LevelDB log, which would have a 
negative impact for request batching, which I fixed [2]
I also found a bug, which might lead to incorrect state machines (while 
the log remains correct) on leader crash and subsequent failover [3].
I ran into an issue today which leads to slow catching up of a restarted 
member when the log size is big [4]. After removing most of the 
synchronization in RAFT, I'll fix this issue and then release a 1.0.7; 
this or early next week.
Cheers
[1] 
https://github.com/belaban/jgroups-raft/issues/119
[2] 
https://github.com/belaban/jgroups-raft/issues/121
[3] 
https://github.com/belaban/jgroups-raft/issues/122
[4] 
https://github.com/belaban/jgroups-raft/issues/123
-- 
Bela Ban | 
http://www.jgroups.org