Regression detected in CSIT

This recent fix patch:

https://git.opendaylight.org/gerrit/#/c/controller/+/87043/

Produced regression in these 2 cluster tests:

1) https://jenkins.opendaylight.org/releng/job/controller-csit-3node-rest-clust-cars-perf-ask-only-sodium/
Cars performance test: 10K cars are configured in a cluster, the leader is isolated from the cluster, when new leader is elected, cars are verified in the new leader. This test fails because not all cars are seen in the new leader.

2) https://jenkins.opendaylight.org/releng/job/controller-csit-1node-akka1-all-sodium/
Upgrade test: A single controller is loaded with old SW (Carbon in the test) and some configuration, after that the node is shutdown and the data (snapshots folder) is transfered to a new SW (Sodium SR2) controller. The new controller fails to boot up and there are GC and OOM errors in the karaf.log.