Imagine a general with a finite amount of artillery shells and a Howitzer on the fritz (i.e. the intended trajectory of the shots are somewhat off the mark today). Given the choice of where to deploy the artillery, the general may choose to concentrate their fire on the narrow beachhead landing instead of upon a widely scattered formation of units approaching across a vast plain.

The artillery that lands on the plain may strike an advancing unit, or it may fall (possibly harmlessly) between a set of advancing units. The artillery that lands on the narrow beachhead is more likely to hit a unit.

This analogy is far from perfect: sometimes mutations are good, which is one primary driver of evolution. Non-coding regions and/or "baggage to be refactored" (paraphrased great-great-gp comment) in DNA (the regions of the plain/beach not occupied by an advancing unit) can absorb "errors". Also, there are other types of mutations (insertions, deletions, ...), aside from the single point mutations that this analogy was attempting to help convey.

The point is: it's like bunching up a lot of important things over a few points of failure. If you increase "the genetic surface area", you lower the chance of the important thing getting hit.

On evolutionary scales, viable DNA has been selected with a lot of non-coding (and sometimes useful) regions, we know that if we reduce that down, we are more likely to be susceptible to fatal mutations on coding regions (e.g. a region that codes for a vital protein).