Cheers Pete!
Catch is, we could not replicate this in the test system yet... We copied the task list, swapped it in all the maintenance items, then changed the header status, which circumvents the problem... but I'm more keen on understanding the root cause and ensuring we don't have to do this again! 
This is the change document log, odd thing is - the change from positive to negative work is the only item for the Operation at that particular point in time, the other is just the last change time
