* test(consensus): prove leader removal can stall
A departing leader can clear peer addresses before its queued commit
notification reaches the surviving voter. Control the removal append
and its acknowledgement to reproduce that ordering without fixed sleeps
or stopping the leader process.
Check the actual notification failure and the survivor's commit index.
Keep follower-removal and three-voter controls, and cover both the default
removal wait and timeout=60.
* test: timeout=60s with follower or leader
Signed-off-by: Anton Antonov <anton.synd.antonov@gmail.com>
* test: fail when leader removal loses its commit
Require the surviving voter to receive the committed removal and remain
operational. Fail explicitly when the commit notification never reaches
its gate instead of treating the stuck cluster as success.
Both two-node leader-removal cases now fail against the existing bug.
The follower-removal and three-node controls pass.
---------
Signed-off-by: Anton Antonov <anton.synd.antonov@gmail.com>
Anton Antonov