Jira ends up with an index that doesn't contain filters, making it impossible to use filters just by providing their name (filter's ID has to be used).
During reindex (both background and foreground), if an exception is thrown during indexing any issue, comment, change history or worklog, Jira will not attempt indexing filters.
If indexing some issue, comment, change history or worklog fails, Jira will still attempt to index filters. Jira might be able to finish indexing all filters correctly or fail indexing just a few of them.
This behavior is a result of first logging ("Reindex All FAILED. ..."), and then rethrowing an exception inside CompositeIndexLifecycleManager#reIndexAll method, making Jira exit a loop responsible for running different indexers.