There are at least 2 articles reasoning why to remove Moo! and this is the first solid article defending Moo! project - I was waiting for one, my obvious upvote for the quality. 

I was hesitating until now, but finally I'm close to my answer. My reasoning is as follows.

There are N keys and one is valid. Moo! is brute testing them one by one. 
The long time it takes  supposedly shows how safe the method is. 

- There was however a chance that proper key was at the very beginning of the set and puzzle would've been solved long ago. It could have been the first tested key, too.  As such the test would be inconclusive and dismissed.

- As we know that there is N keys and only one is valid, to test how long it would take for a computer of type T to find it, we can just run a quick test, measure the average time to test one key and using basic math we can calculate mean time and maximum time required. Process can be repeated each year within one day on the new generation computer.

My conclusion is: why to run hundreds (or thousands) of computers for many years, if we can use one computer for one hour once a year and achieve the same or even better result?

I can add all natural numbers from 1 to 1 000 000 on a piece of paper  within seconds (using a clever method) instead of weeks or years and tons of paper (adding each number one by one).

If there is a better and much quicker method to solve the problem we should use it instead of brute-force method.