Today I got my feature done in an hour, but I spent 3 hours wrestling Spring, JBOSS, and Eclipse.
I would just like to restate some of my principles now.
* Tool performance and stability is important.
* Iteration time is important.
* A good error message can save an hour of time.
Which is to say, if you pay someone $40 an hour, you should buy them a $400 tool if it saves them 10 hours of time over it's lifetime.