First identified with the failures on the `InitialClusterSizeTest` test [1]. The test initializes multiple nodes concurrently. During the cache manager creation, the follower nodes retrieve the global rebalance status from the coordinator in a blocking manner. If a view change happens during this retrieval, the initialization hangs. The request eventually times out, and everything proceeds, but the view update seems not to be applied after. Since the follower has an older view, joining the caches with the coordinator will also hang.

This test also has failures on the 14.0.x branch.

[1] https://ci.infinispan.org/job/Infinispan/job/PR-11672/16/testReport/junit/org.infinispan.remoting.transport/InitialClusterSizeTest/testInitialClusterSize__Flaky_Test_/