ORIGINAL: siclick33
Having thought about it after HarryC's explanation I can see why it would work. I didn't realise that the exhaust velocity in our little engines would be supersonic and therefore the duct would be working 'the other way round'.
It also works when the flow is subsonic and slows down in the divergent pipe. Although the flow speed falls its pressure rises and the total multiple of gas energy and pipe area is greater than the same sum at the start of the pipe due to the area rising more rapidly than the gas energy reduces. Thus the force at the back of the divergence is greater than at the front of the divergence and the result is a net forward force on the pipe.