The following material is flagged Green Level. It is intended to reflect material that the author believes to be a matter of consensus among experts in the field. This belief may be incorrect, however; and as the author is not an expert and does not have an expert fact-checking the article, errors may creep in.
Let us assume that we have a population with mutants. Continuing from the last example, let's say we have bears, some of which have coats better able to trap heat. Then, let us say that this mutation becomes universal; in the last example, all of the bears become better able to trap heat.
Now, let us assume that a second mutation shows up. This mutation repurposes the protein responsible for fur that grants resistance to cold so that when expressed in fur, it will still have its original purpose, but when expressed in blood, it will grant further cold resistance by the same amount. (Yes, I know that this specific for-instance has an incredibly low chance of actually making any kind of biological sense. But we're discussing hypothetical abstractions anyway, and it gets the point across.) So, the two mutations spread through the population at about the same rate, right?
Wrong. Remember that the second mutation works by repurposing the first. In bears that do not have heat-trapping fur, the mutation for cold-resisting blood will, at best, do nothing. At worst, it will repurpose some other protein, and do something very interesting to the bear's internal biochemistry. So, the cold-resisting-blood mutation can at best spread through the population at the same rate as one with a neutral effect, and at worst cannot spread beyond the subset of the population that has the heat-trapping-fur mutation. In other words, the mutation is of no beneficial effect without the one it modifies.
Next, suppose that, once the cold-resisting-blood mutation has become universal (and of course, the heat-trapping-fur mutation with it), a mutation happens in the heat-trapping-fur gene that causes it to function better, but only if the cold-resisting-blood gene is also present. (And at this point, the specific examples start approaching impossibility, but again: for-instances, hypothetical abstractions, gets the point across, et cetera.) Since the cold-resisting-blood gene is universal, the heat-trapping-fur-mark-2 mutation has no drawbacks. Since a heat-trapping-fur gene still exists (and can still be adapted for the purposes for which the cold-resisting-blood gene uses it), the system still works. And so, we have a system that is stable as it is, but which will fall apart if any component is removed.