In addition to the higher unit count there are a few bullet points in the article that indicate where saved/additional time is spent. From the sound of it, most importantly: "The DE pather was modified to have a much higher max iteration count than Age1's, so longer and more complex routes can be found."

In addition to the other replies, depending on how the game is written, it can also help with battery life if the speedup allows the CPU to go to sleep (or lower it's frequency) more often.

They also increased the iteration time to search for more complex paths, so I’m sure this helps with that.

IIRC the max unit cap is much higher.

The fact that you don't need to optimize something doesn't mean you shouldn't.

It does, actually. At least don't do it on my dime. Obviously learning is another ballgame.

At the least, it implies that you ought to carefully consider whether you should. Optimized code often makes trade-offs in complexity, flexibility, etc, not to mention the extra cost in development time.

With those costs, there should be a compelling benefit that you get out of it. Otherwise, it's not worthwhile.

Avoiding premature optimization is one of the most important best practices of modern software development.

This isn't premature... the algorithm and performance are understood. That said, I agree that it's not always the best use of time.