Optimizing index during intensive work causes problems

Sometimes JIRA writes in log:
INFO [jira.issue.index.DefaultIndexManager] Optimising the index because 4000 issues have been re-indexed since
last optimise. Threshold is 4000.
After that it's locks indexes and overall performance. It takes about 1,5 minutes, during which we have average 5-6 updates (sometimes 10-15). All this updates are finished with IndexException. The result is outdated indexes for all failed operations (Issue navigator values don't corresponds to real state of Issues)