There are two reasons why I am marking this as RFC as of now. I don't see the preparatory patches yet, I wonder if this should be based on for-next-20190415? I couldn't reproduce the issue, which theoretically should exist as in the patch 2/2. Some more cleanups related to the inode flags such as [1] [1] (don't change binode->flags directly, but do all flag updates on a temporary variable) is wip.
