in cloud, since iops and bandwidth limitation, write operator in raft is slow, so raft state (logterm, lastlog, commitlog...) are the same among the peers LeaderCandidates select all peers can be select as a leader during the restore.

RecoverData recover the tikv cluster 1.

for region A, has candidate leader x, y, z peer x on store 1 with storeBalanceScore 3 peer y on store 3 with storeBalanceScore 2 peer z on store 4 with storeBalanceScore 1 result: peer z will be select as leader on store 4.

for test.