Zookeeper throws timeout after several restarts of nodes

I have a problem with zookeeper deployed on kubernetes using Confluentinc Docker Images.

### Cluster setup: ###
- 3 nodes in digital ocean ( 4 CPU, 8 GB or RAM) as kafka dedicated nodepool. ( Only zookeeper and kafka are running here, and some monitoring daemon set).

- If a zookeeper node/pod ( any node) restarts several times the zookeeper cluster is not accessible anymore, and to make it work I have to redeploy all the nodes.

- Once the node rejoined the cluster, all zookeeper nodes start to show this kind of error:
Each node has one kafka pod and one zookeeper pod.

### Errors one each node: ###
`zookeeper-node-0 logs:`
[2020-10-29 00:29:20,852] INFO Refusing session request for client /10.244.4.86:47054 as it has seen zxid 0xd00000000 our last zxid is 0xc00000004 client must try another server (org.apache.zookeeper.server.ZooKeeperServer)
`FULL LOGS:` [node-0-log.log](https://github.com/confluentinc/common/files/5465751/node-0-log.log)

`zookeeper-node-1 logs:`
244.3.185:36040 as it has seen zxid 0xd00000000 our last zxid is 0xc00000004 client must try another server (org.apache.zookeeper.server.ZooKeeperServer).
`FULL LOGS:` [node-1-log.log](https://github.com/confluentinc/common/files/5465752/node-1-log.log)

`zookeeper-node-2 logs:`
[2020-10-29 00:28:59,554] WARN Unable to read additional data from client sessionid 0x102508ff4380002, likely client has closed socket (org.apache.zookeeper.server.NIOServerCnxn)
FULL LOGS: [node-2-log.log](https://github.com/confluentinc/common/files/5465754/node-2-log.log)

although the zookeeper cluster seems to be up, I'm not able to get anything from it. If I try to use cli to list a path I'm getting a timeout.
Any Ideas?

If this is not the place for this issue please guide me.
thanks

Start with the linked node-0, node-1, and node-2 logs, focusing on the zxid mismatch and timeout messages after repeated restarts. Review the described Kubernetes deployment and Confluent Docker image configuration to determine why a rejoined node cannot serve sessions. Done means identifying a reproducible cause and documenting a verified recovery or configuration change.

Written by the indexing model from the issue text.