Apache Ignite Rebalancing

I was working on Apache ignitev2.17 and started upgrading to v3.0 . I have 3 server nodes and .net api as a client. I created one zone with Replicas 2 and scaledown = 5 and created a table belong to this zone, then I made one of these nodes offline. the cluster keeps working fine (I can check cluster status, zones,..) but if I do query to data, I got issue like this

```
SELECT * from t_read_status limit 1;
SQL query execution error
Mandatory nodes was excluded from mapping: [Node3]
```

or

```
select count(*) from t_read_status;
SQL query execution error
The primary replica await timed out [replicationGroupId=17_part_1, referenceTimestamp=HybridTimestamp [physical=2025-10-28 11:47:20:016 +0000, logical=0, composite=115451628094488576], currentLease=Lease [leaseholder=Node2, leaseholderId=486d7bc8-b03e-4c0f-91a8-76a6bb0dc243, accepted=false, startTime=HybridTimestamp [physical=2025-10-28 11:47:14:874 +0000, logical=95, composite=115451627757502559], expirationTime=HybridTimestamp [physical=2025-10-28 11:49:14:874 +0000, logical=0, composite=115451635621822464], prolongable=true, proposedCandidate=null, replicationGroupId=17_part_1]]
```
and I checked the online node logs and found
```
[WARNING][%Node1%replica-4][ReplicaManager] Failed to process the lease granted message
```

then I tried to change the cluster config of replication.leaseExpirationInterval to 30000 then I did the test again I have the same result but when I check the node logs found
```
RebalanceRaftGroupEventsListener] Going to retry rebalance [attemptNo=16908, partId=20_part_18]
[ERROR][%Node1%JRaft-Request-Processor-8][ReplicatorGroupImpl] Fail to check replicator connection to peer=Node2, replicatorType=Follower.
[RebalanceRaftGroupEventsListener] Number of retries for rebalance exceeded the threshold [partId=17_part_18, threshold=10]
```

I mentioned some of the logs , and I am lost don't know why the cluster (raft) can't do rebalancing

Start by reproducing the three-node setup with replication.leaseExpirationInterval changed, then inspect the RebalanceRaftGroupEventsListener, ReplicaManager, and ReplicatorGroupImpl log entries cited in the report. Done means identifying why rebalancing fails after Node3 goes offline and documenting or correcting the behavior so queries continue working with the configured replica count.

Written by the indexing model from the issue text.