If a ranking operation fails at the database level (when Active Objects attempt to persist it into the database) it's possible that the instance can get into a state where duplicate log entries exist. This appears to cause ranking to stop updating and means JIRA needs to be restarted to fix the problem.

Whilst this workaround can work OK for smaller companies it affects Enterprise organisations significantly. Can you please enhance JIRA Agile to more gracefully handle this in-product so it doesn't need a workaround?

The workaround is covered in full detail in How To Fix Issue Ranking Operation log entry found where none expected.