HDFS-5138 was such a total screw-up. I'm astounded as to how many people were watching that JIRA and still thought it was OK to completely nuke operational procedures out of the water in the non-HA case.
I've had discussions with Sanjay Radia to restore -finalize so that future generations of branch-2 users aren't caught by surprise. At this point, especially with branch-2 regularly tossing operations folks under the bus if it is worth the effort. It might be easier to go ahead and remove the -finalize option so that it's a fatal error rather than silently failing like it does now. Charles Wimmer, do you have any opinions on this?