[ https://issues.jboss.org/browse/JGRP-2092?page=com.atlassian.jira.plugin.... ] Neal Dillman commented on JGRP-2092: ------------------------------------ This scenario happened several times in our lab. We no longer see the issue as it was worked around with staggering responses for all of the "bulk response" messages (ie: view change, etc). However, I believe that this could happen again with a marge enough network and low enough bandwidth. In other words my ounce of prevention is working out, be the pound of cure would still be the best solution. I had created a unit test as well and hound the same results. We also ran into partial scenarios where the consistency checker was being run by one functioning coordinator, but other, non-functioning coordinators would not respond to the merge and/or the coordinator running the checker did not think it should be merge leader. In all cases the issue is complicated by the fact that none of the non-coordinators will trust a new coordinator that is not a member of their group. I believe that a solution will involve the 2nd rank host (and maybe the 3rd?) running a checker to see if the coordinator's viewInfo shows the coordinator as thinking it is, in fact, a coordinator. -- This message was sent by Atlassian JIRA (v6.4.11#64026)