Properly staggered and with a lead time of 9 months, yes, 222,375 women could deliver a child every 0.000040472 months on average. I'm sorry, was that not the question? :)
[1] basically from scratch; it's suitable for smart highschoolers with some programming experience. That website has been responsible for more than a few IOI medals.
I remember doing some of Dave Thomas' code katas when he first posted them, and thinking that they were fun but a bit pointless, and generally rather too easy to learn much from. The other month at work I did a presentation at work on Andy Hunt's theories of programmer skill development where I mentioned katas, saying that I didn't really see they had much value. However, Ron Jeffries' celebrated failure to write a sudoko solver puts a slightly different perspective on this.
If your daily work involves doing algorithmic stuff, even if it's only moderately demanding, then I don't see that you will get much benefit from doing the same sort of thing at a more noddyish level.
However, if your daily work isn't like that, if perhaps it is in the vicinity of what used to called Data Processing (e.g. implementing payroll systems), then I can easily imagine that your coding chops might get rusty. In which case, taking deliberate steps to avoid that would be sensible.
[This argument only applies to the question of puzzles improving ones work, rather than being fun.]
If you're at work, no, it's not about clever problem solving. In my spare time though, I prefer to solve bigger problems than parsing CSV files into databases.
