Consider the following scenario. (This is exactly what happened to us and not a contrived scenario)

JIRA is installed in a production location and everything is just peachy.

A new release of JIRA is available and so it is installed in a sandbox area for test driving purposes. Data for this sandbox instance is obtained from the aforementioned production instance, i.e. you export from production and import into the sandbox.

A big error has crept in!

The data in the JIRA tables includes the location of the backups, attachments and indices. So, unwittingly, the sandbox instance is pointing at the same location as production for backups, attachments and indexes.

You notice this and proceed to update the index location for the sandbox instance to a sandbox location.

Gradually the users are increasingly insistent that JIRA (on production) isn't working. Filters are throwing exceptions and the like.

You start piecing together apparently disconnected information and deduce the cause:
The reason for this is that when you changed the location of the index in the sandbox instance, JIRA-sandbox first blew away the indexes at the old location before creating it at the new location. AAAARRRRGGGHHH!

Now, the title of this issue should make sense.