Okay, correct me if I'm wrong (as Patrick always helpfully does), but there seems to be a few points that people keep missing:
1. This is simplistic and slightly specious, but basically the speed of the CPU is what Intel says it is. PERIOD. As Patrick points out, this is based on extensive testing and statistical analysis. It makes absolutely no difference if the chip is identical in every way to others rated at higher or lower speeds and will run (potentially only in the short term) at a different speed at stock or different voltages. Intel certifies and locks the chip at "X" MHz and that is the only Intel GUARANTEED and acknowledged speed. It's sort of a semantic argument, but since they are the manufacturer, they have the right to determine and guarantee the speed regardless of others' assertions that there is more "under the hood". They also have the right to price the chip as they see fit.
2. Silicon is a substance with natural variations in its structure, even under the most exacting manufacturing processes. I'm an architect and, this is a crude comparison, but, whenever one specifies any product that is subject to natural variation you receive what is called "range samples". Take granite for instance. You specify a certain "species" (this is a bit of a misnomer), finish (polished, honed, flamed, etc), and other criteria. You then receive 6 or 8 finished samples for review that are representative of the range of that stone found in that particular quarry (analagous to the silicon wafer?). As you can imagine, these can be dramatically different from each other, and you may reject some as being too far "outside" the approved range, while others are better in appearance than your original sample. Grades of framing lumber is another example that many are familiar with. Out of the same tree you can get a very wide variation in structural strength, dimensional stability, grain direction, etc. Almost all of the wood is tested and used, but some boards are certified to meet more stringent structural standards than others, even if only marginally so. This largely has to do with how the log is sliced at the mill, which area of the log the board came from (analagous to the die's position on the wafer?), how it was kiln dried (burned in?), etc. Could you use a lesser grade of lumber to frame the same building? Structurally speaking, probably yes since the grading has safety factors built in. Would you want to take-on the liability that the structure is sound (for MANY years) WITHOUT the manufacturer guaranteeing the parts you built it from are? I wouldn't.
3. Obviously, this is why certain grades/pieces of granite, wood, whatever, are priced differently even though they came from the exact same quarry strata, tree, what have you, and required the same amount of manufacturing effort to produce. People accept this kind of price/quality pointing as natural in other industries, but for some reason with microprocessors there's the idea, particularly amongst the overclocking crowd (of which I'm a member) that there's excessive price gouging going on. If you don't like the price, don't buy the product, but believe that producing, testing and certifying is an expensive process.
4. There is simply no way of checking the veracity of the "600E is really an 800EB" argument without either testing a group of Intel certified 600Es under the exact same controlled conditions that the 800EBs were certified and getting the exact same (statistically speaking) pass rate, OR running the 600Es at 800 at the Intel spec 800EB voltage for SEVEN years and seeing if they live up to the same standards that the 800EBs are guaranteed to (The first test is probably impossible as the testing conditions are pre-packaging, etc.). Simply saying "600Es have the same multiplier as 800EBs, and many run at 133 FSB/800MHz: Therefore 600E = 800EB" IS a specious argument.
Sorry if this is long winded, preachy, over simplistic, obvious or if I simply missed the point .