I'm wondering if I should really report this as work has been done on MERGE3 but I detected a weird case where one node is present in many subgroups. First of all I wanted to know if this was normal (i.e. could happen) or if it is again related to my use of FLUSH when installing views. I've replaced the replica names with {A,B,C} this time for brevity and what I'm seeing is this:

2011-11-24 10:28:52.236 [FINE] - MERGE2: A found different views : [A|3], [A|4], [A|2]; sending up MERGE event with merge participants [C, A, B].
Discovery results:
[B]: coord=A
[C]: coord=A
[A]: coord=A
... (more logs)
2011-11-24 10:28:52.376 [FINE] - FLUSH: A: installing view MergeView::[B|5] [B, A, C], subgroups=[[A|3] [C], [A|4] [A], [A|2] [B]]

If this is a possible occurrence then I need to again rethink my state merging algorithm since I only delegate the task to a former subgroup coordinator. But in this case only one node is ever concerned. FYI, my cluster is composed of 3 nodes, all under VMs, and the bug happenned when the host machine of B and C started hanging.

I can attach log (with DEBUG level) if needed