In single-instance mode, during re-indexing users aren't able to access Jira web interface: they are redirected to a page saying that Jira is currently blocked because of a re-indexing.
With Scarlet, users are still able to access Jira on a different cluster node from the one that is actually executing the re-index operation, corrupting so the re-index operation by creating/modifying issues.
This is clearly wrong, so during re-indexing users on all nodes must be always redirected to the "blocking" page.