The OnDemand Importer is currently too lenient and allows XML imports to contain issues that have

• same issue num ("2" in the following example)
• same project ID ("10000" in the following example)

  id   |  pkey   | project |        created         
-------+---------+---------+------------------------
 10002 | ICNET-2 |   10000 | 2011-03-22 21:20:36+11
 10040 | BZI-2   |   10000 | 2011-03-16 22:16:48+11

That inconsistency will cause the upgrade task introduced in JIRA 6.1-OD-3 to fail (see this JRADEV for more details), and will prevent JIRA from starting up until the database is manually fixed by Support.

The expected behaviour if for the importer to fail fast when such data is detected in the import file.