...I have yet to see the makings of a grand conspiracy, rather than the petty bullying of the powerful over the weak, the insider of the outsider. I'll take the statements of this particular group of scientists with a little more salt in the future. But as far as I can tell, the weight of the evidence--and what we know about the history of the planet, and carbon dioxide--still seems to be on their side.
But here's the problem: what is the "weight of the evidence"? This group's original data set was destroyed. Inadvertantly? Possibly... but considering that they were discussing destroying it in the event of a freedom-of-information request, that too is questionable. As one of McArdle's commenters said, "One notes that, in both law and common sense, when a person destroys evidence, you are allowed to presume that it supported his opponent's case."
Sure, it's not the only evidence. But there have been problems with every data set I've heard of: new urban heat-sinks where temperature stations used to be rural, cherry-picked tree ring data, glaciers growing where not measured for an AGW paper, shrinking where the AGW boosters measured, inability to explain the mechanism behind either an ice age or a warming period, pre-1960s temperature data that dramatically smooths out temperature variance graphs, Mars warming up in the absence of coal-fired power plants and SUVs... With all this uncertainty out there - uncertainty noticed only by "skeptics" (another McArdle commenter noted that skepticism used to be considered a virtue in science, not a vice) - one could be forgiven for believing that the really bad acting on the part of these CRU folks calls their findings into question.
I used to be a geologist. In the summer before I graduated, I worked for a gold mining company, collecting rock samples from a very steep hunk of the Sierra Nevada. Here's how you collect a rock sample: you locate yourself on a map, find a rock where you're standing, whack it with a rock hammer until you can collect enough pieces of it to fill a bag about the size of a quart of milk. Mark the location on your map with a unique identifier, mark the bag the same way, stick the bag in your backpack, take it to a lab. The first step - locate yourself on a map - is VITALLY important, because if you don't know where your sample came from, you can't draw any conclusions about where the most gold is. (Samples A, B, and C, in an east-west line, show increasing gold concentrations to the west; perhaps you could put a mine toward the western extent of your sampling area; you certainly wouldn't put it to the east. But say you mislabel your bag - I did that ONCE - or you can't find yourself on the map; where do you dig?)
My map, the best one available for my field area, had 40-foot contours. Think of contour lines as the "bathtub ring" around an object - put a big irregular rock in a tub, fill the tub exactly an inch deep, and look down on it from above. The "bathtub ring" around the rock would be exactly level at one inch, and would show the shape of that rock at a height above ground (the bottom of the tub) of one inch. Now fill the tub two inches high and look down on THAT bathtub ring to see the shape of the rock at two inches above ground. Et cetera. Draw those bathtub rings and you have a contour map of the rock, with bathtub rings closer together where the rock's side is steep, farther apart where it's more gently slanted down. Now scale up: a field area five miles square or so, with the "bathtub rings" at 40-foot intervals. It's mountainous - but most of the outcrops of rock, ridges, cliffs, are shorter than 40 feet. So the map, friends, is nearly useless: you can't tell where you are from it. I could be standing next to a thirty-foot ridge of rock as steep as the side of a building, and it'd be nowhere on my map if it happened to fall between, say, the 5,000- and 5,040-foot contours.
So I had to start from a known point (say, the end of the high-center dirt road I drove to get to the area), tie off a 200-foot nylon tape measure, take a bearing with my trusty compass, and walk to the end of my tape. Mark that spot, on the ground and on the map. Go back and untie my measure, return to new known spot. Tie off measure, take a bearing, walk (or sometimes, essentially rappel, using my tape measure as a line - though the steep slopes naturally mess up the horizontal measurement)... until I could get to the area from which I was to collect a sample. Oh, and sketch in those 20- and 30-foot cliffs and ridges, for the next schmuck. Took a long time.
The point is, my model - my map - was missing a LOT of data, and in fact bore next to no resemblance to reality. I had to fill in those data as best I could, and I had to do it with an eye to the next baby geologist who would be sampling there - he or she would need to know where things were, just as I had. So I had to be (a) careful and (b) transparent. But even so, my map at the end of summer was only a little better than the one I started with, because it only improved where I was sampling; to look at my map, the next baby geologist might think, "Wow, this area here where Jamie was working sure had a lot of cliffs and ridges; luckily it evens out over here in the part she didn't get to. I'll start there!" And that baby geologist would find him- or herself rappelling down scree slopes hanging onto a 200-foot tape for dear life, just as I had.
The dangers of relying on a model, that's what I'm talking about. And when you throw in incomplete reality checks, reality checks in only the convenient or hypothesis-confirming places, and deletion of reality checks that either do not or might not agree with your model, well, the best face you can put on that is that, poor you, you're going to get all wrapped up in the model and lose the ground truth, ending up looking like a fool; the worst face is that you're trying deliberately to deceive.