For a number of years researchers have been trolling the genomes of numerous species, looking for conserved or 'ultraconserved' DNA sequences (ultraconserved sequences are stretches of DNA that are the same in mice, rats, and humans and perhaps other species, and often similar in fish), assuming that they represent regions with important regulatory or coding functions which is why they've been maintained by natural selection. The idea is that these stretches of DNA are so crucial that without them, the organism would die. However, often when these regions are knocked out (experimentally deactivated), there is no effect on the animal. This has surprised many people, given the well-accepted equation of conservation=importance.
From an evolutionary perspective, the problem is that mutations are always occurring and no bit of DNA is invulnerable. Given that, over time every nucleotide will experience mutational hits; most of the new alleles may disappear by drift, but not all of them will. Eventually, there will be no recognizable sequence left (that is, if the corresponding sequence in distant descendants could be identified, they would bear no similarity). Selection can maintain sequence if it is functional, and if therefore most mutational changes are harmful. But otherwise, other than by unlikely aspects of chance, how can deep conservation occur?
The conservation=importance equation thus makes sense and fits evolutionary theory well--but unfortunately for evolutionary biologists, nature doesn't always follow the rules. There are several possible explanations for this. Linear DNA sequence is less important in many contexts than the three dimensional conformation a stretch of DNA folds up into in the cell. It's this three dimensional shape that determines what other stretches of DNA or proteins can bind to it, and thus its function(s), and there might be multiple ways to attain the same shape.
And, sometimes it's not the shape of the molecule that's so important but some other characteristic such as, for example, its acidity, which is determined by amino acid composition, and acidity determines what that stretch of DNA can bind to. There are many nucleotide and amino acid combinations that will produce the same acidity. That means that the specific DNA sequence may not be conserved, but other characteristics of that stretch of DNA may be.
Many gene knock-out experiments in mice, even if the gene isn't one that is ultraconserved, have shown no effect on the mouse. Even large regions with presumably essential genes or regulatory sites have been knocked out and the mice seem none the worse for wear. This is perplexing, except that the mice are only observed in ideal conditions in a laboratory setting, while the missing DNA may be important when the mice are in other contexts.
Finally, whatever works is what nature does, and whatever works in a given context can and does vary widely. There are no steadfast rules for how to get from here to there, and there are exceptions to every generalization. Natural selection isn't always the explanation, nor do specific genes and their evolutionary histories work the same way in every situation, and so on.
To us, this story is a reminder that any rule an evolutionary biologist can come up with, nature can break.