If a shard is undergoing a long-running force merge and we decide to relocate it onto a different node, I believe (but have not checked) that we do not take any action to cancel the force-merge. Is that right? Do we do the right thing in terms of tracking the disk usage of the shard while the force merge is still running?