We have had incidents where it's Master/Slave replication continues on failure when configured to shutdown the master and slave brokers. It doesn't achieve the consistency it lists as a feature, and provides zero visibility of progress. I feel many of the features the project lists are incomplete, and are listed as a grasp at straws to have the bigger list.
Due to the poor quality of the replication, we had to invest a tonne of time to implement replication at the block device level with DRBD+GFS2, complicating the system and (benchmarks pending) likely decreasing performance, all to achieve a 'feature' ActiveMQ boasts.
I would also challenge a project to have worse, more incomplete documentation.
At the same time, I thank the ActiveMQ team for their contributions and ask the community to not turn it's back on this potentially decent solution. The ideas are solid. It isn't production ready, I feel, but that just means it needs some love.
Cheers,
Tim Vaillancourt