We are currently having the same issue. One of our nodes is basically about to crash again due to this issue. We have had this issue happen at least 3 times before in the last 2 years. Our solution has been bring up a new cluster, let the data and indexing fill up the new cluster and then take the old cluster off line. We have never felt comfortable enough to just failover and rebalance due to various issues in the past.
We are on Community 6.6.0 and cant go to any higher version of 6.6 because they dont have anything we can download for it. We cant go to 7 because they got rid of being able to do XDCR and XDCR is our one of our saving graces for not doing a failover on a live cluster.
Has anyone been able to fix this issue in community without going to version 7?