I completely agree with this. I fact when I have a bigger architectural problem to think about I like to sit on it for a day or two, thinking about one or two designs that would work. It takes a while to see the strengths / flaws in each design and jumping in to code you won't realize problems until you have something half implemented.