Retrospective insights from my experiences in setting up a lab at UCSD, in random order (I'll keep adding to it over time, they are in no real order, some are just generic things):

Added 8/15/13: MIND THE P'S. There are only three things that are important. The three P's - People, Projects and Pennies. 

1. Spend more time in the lab than in your office, especially in the first three years.
2. Never put all your eggs in one basket (unless you have unrestricted funding- HHMI, Janelia, etc).
3. The people working with you are your most valuable resource. Choose wisely and cherish them.
4. Focus on the Science, not on the money. No matter how many $$ you have tonight, one day it will be zero. No-one (including yourself) will remember the big grant you got in 1999. Everyone will likely remember the cool paper.
5. Have opinions, but be ready to change them if needed. Very few PI's are wilting violets.
7. Don't enter pissing contests, but if someone wants one, don't walk away. It will be seen as weakness.
8. Collaborations only work if: a) there is something very little that you want from someone who is doing that anyway or b) someone is extremely interested in the same questions that you are asking (rare). They are almost always useful only when the collaborator does something that you can't. Expect only ~ 50% of collaborations to be really productive, learn to prematurely stop the ones that are going nowhere, early on in the process. 
9. START small and preferably STAY small. 1 tech over the first 2 years (+you working) is more than enough to get a solid publication. The question is not what you can afford, the question is what is the right move.