Confluence data center has slow memory leak

https://pug.jira-dev.com/wiki/pages/viewpage.action?pageId=1492615395

Steps to reproduce:

• Setup two node cluster and set cache size of `Page Ancestors` to 500 (to make overflow easier to reproduce)
• Stop both nodes
• Start node 1
• Use `./managePages.py` from https://bitbucket.org/akdominguez/populate-confluence-scripts. Option 1 (tree) with 100x100 dimension
• Start node 2
• Observe cache to go back to 80% full when close to 100% on node 1
• Observe cache to be over 100% eventually on node 2

Github issue opened: https://github.com/hazelcast/hazelcast/issues/5485

Release process: https://extranet.atlassian.com/display/DEV/Hazelcast+3.4.2

PR: https://stash.atlassian.com/projects/CONF/repos/confluence/pull-requests/5972/overview
Branch View: https://collaboration-bamboo.internal.atlassian.com/branchinator/69271566/confluence_58%20(read-only)/issue-58%252FCONFDEV-34149