Running force merge against a read-write index (an index that can still have read/write activity post force-merge) can cause very large segments to be produced (>5Gb per segment), and the merge policy will never consider it for merging again until it has 75% of deleted docs. This can cause very large segments to remain in the shards.

This is not currently documented (https://www.elastic.co/guide/en/elasticsearch/reference/6.2/indices-forcemerge.html). Would like to see a warning message added to the above page to specify force merge as an API call to use only against read-only indices.