A worst case performance of O(∞) might not even make it an algorithm according to some.
An algorithm is just a series of steps and you can always do worse by tweaking it a little bit to get the desired output in more steps than it was previously taking. One could purposely put the knowledge of the number of steps taken into the algorithm and make it terminate and produce the correct output only after X number of steps have been done. That X could very well be of the order of O(n2) or O(nn!) or whatever the algorithm desired to do. That would effectively increase its best-case as well as average case bounds.
But your worst-case scenario cannot be topped :)
