The more closed the position the longer castling can be delayed, generally speaking.
In some well-analyzed openings a player may routinely not castle at all. If memory serves, Anand did so in the semi-slavs of his match with Kramnik. Someone correct me if I'm in error here.
Not an example for the average player though.
castling early automatically is good when youre starting out. but assessing each position idividually and seeing which rule(s) are strongest (to help form a plan) has to come as some point.
really analyse Stienz imortal game. the king seems so save in the center for a little while longer but then gets ripped to shreds.
I posted an article on my website discussing this topic
http://www.cob.sjsu.edu/splane_m/chess/delayedcastling.htm
Thanks!
Some good thoughts at your link Spiane.
I seem to be going through a growing phase with this also JohnOh. Like you stated, I would always go for the middle, get the K tucked away, and then work out how to attack. Recently however, my coach has scolded me a time or two for castling too early in games. As a guide line, He says if possible, TRY to wait until your opponent castles first. He stated this is not a hard fast rule, but basically you don't want to give away too much information too soon if you don't have to. He also pointed out some interesting things regarding the enemy's piece deployment. Like if white is developing his pieces aggressively aiming everything at the King side, or pushing pawns on that side prior to your 0-0, you may want to hold off castling over there, and instead, make it possible to go either way, or perhaps even stay in the middle and close things off. Basically I think it boils down to assessing the ebb and flow of each game and making the correct judgment for each. Kind of leaves it up in the air huh? lol Maybe another way of looking at it is, stay flexible as long as possible. You can't undo castling, but by waiting to do it, you at least leave yourself two other options for a while if they become necessary.
I do find that I really have to make a concerted effort to look for other useful things to do, instead of my old automatic 0-0 after I've achieved my opening goals. It's amazing how habit forming some ideas can be in chess as we learn, and how hard they can be to break after learning them.
Regards!
Also, if the queens leave the board early it's more possible to delay castling. The idea is that without the queens there are fewer mating threats. In that case the king might be better positioned in the center.
I'm a notorious late castler myself.
But I have this thing about pushing h pawns supported by a Rook.
One day, everybody will play this way 
In the past, I've always had an idea of a three-phase opening: 1) going for the center, 2) castling, and 3) preparing an attack. However, recently in my studies I've seen a couple GM games where one player delayed castling the king until sometime in middlegame.
Are there any general guidelines as to when this is appropriate?