"./manage.py test" is your friend. Deploy the update on a separate server (or your development box) and run every conceivable test. Update your test fixtures with data that resembles production data and use it for testing. Do stress testing on a regular basis and monitor your machines (I love Munin for that) closely.

Also, remember to make tests that are performance-dependent, so any relevant change is flagged (you can kind of self-scale your test timeouts by using "getting item A does not take more than twice as long as getting item B").