For testing this you need two different sites where you run upgrade. Test 1. Upgrade from 3.0 In the 3.0 site create feedback module on the front page, add some questions Add feedback block to two courses As a student complete feedback in both courses by clicking on the feedback name in the block As a manager make sure that you can view feedback responses filtering them by course Upgrade the site As a manager make sure that you can view feedback responses filtering them by course Test 2. Upgrade from 3.1beta In the 3.1beta site create feedback module anywhere, add some questions Complete the feedback Go to the database, open table mdl_feedback_value and duplicate one of the values If you look at analysis or responses for this feedback now there will be some mess and debugging messages, ignore that, this situation hopefully can never happen IRL Hack mod/feedback/db/upgradelib.php and add some output in the beginning of mod_feedback_upgrade_courseid() Upgrade the site Make sure the hacked output was not displayed during upgrade (which means that this function was never called) Look at analysis and responses page for the feedback, it should be consistent now as duplicates should have been removed during upgrade