Good point. I think also it would find the optimal (in the worst case time) if it continued to search for shorter solutions over the entire space once it found one. If it happened to find a solution of length 14, for instance, it could stop considering those of 15 or more, so you would still avoid the 2 month waiting time (same can be done with DFS). Though come to think of it you probably mean to do this with iterative deepening by not throwing away the intermediate structure (like BFS).
What I really appreciate about this article though is the solid presentation of math connected with computer science and the author throwing around multiple programming languages. They don't even see the code anymore...