In practice, I suspect that we have numerous ways of dealing with mutations. However, if I recall my high-school biology correctly, during DNA replication there is error correction. My guess would be that evolution would tend to regulate mutations rates through the efficiency of error correction mechanisms (and the accuracy of the primary duplication mechanism) more so than a meta-evolution in the structure of DNA. Having said that, it is reasonable that there would be evolutionary pressure to have DNA that requires more mutations before becoming dangerous.

Responding to the content of your post. DNA is not executable code as much as data. The 'primary' purpose of DNA is to provide the blueprints for proteins. DNA sequences can be thought of in by grouping the base pairs in 3 called codons (this grouping is not reflected in the molecular structure). The beginning and end of a gene are marked by "start" and "stop" codons, which are just a specific DNA subsequence. Conceptually, a gene is converted into a protein by replacing each codon with its corresponding ammino acid (based on a lookup table) [1]. The function of a gene is determined by the chemical properties of the resulting protein.

As it turns out, there are parts of DNA that do not code for proteins, yet still have significant effect. However, I am not aware of any evidence of a naturally occurring use of DNA that resembles executable code.

[1]http://www.biogem.org/codon.jpg