It can be beneficial for the ConcurrentMergeScheduler to sometimes not do concurrent merges (i.e. sometimes do a serial merge). When the provided "OneMerge" is small and when the MergeTrigger is FULL_FLUSH (i.e. on a commit), a new searcher would benefit from seeing fewer segments. If index replication is used, this setting can reduce the net replication by merging a little more eagerly sometimes.
Legacy Jira details
LUCENE-8965 by David Smiley (@dsmiley) on Sep 04 2019, resolved Jun 21 2020
Linked issues:
It can be beneficial for the ConcurrentMergeScheduler to sometimes not do concurrent merges (i.e. sometimes do a serial merge). When the provided "OneMerge" is small and when the MergeTrigger is FULL_FLUSH (i.e. on a commit), a new searcher would benefit from seeing fewer segments. If index replication is used, this setting can reduce the net replication by merging a little more eagerly sometimes.
Legacy Jira details
LUCENE-8965 by David Smiley (@dsmiley) on Sep 04 2019, resolved Jun 21 2020 Linked issues: