There's no way for 'natural selection' to "favour" an heterozygous population. Unless this is some sort of selection goddess, seeing the undesired genes in spite of their recessive state., that is.
Like I said, the context of the what they are talking about, they are referring to the distribution of alleles in a population, not necessarily the number of alleles in the population (if that makes sense).
For example, say I had a gene called "Gene A" with two variations, A1 and A2. If a population is 90% homozygous for A1 and the rest of the population is heterozygous for both genes, this means 95% of the alleles in the population are A1 and the remaining 5% are A2. So there is low variability in the population because most versions of gene A are the A1 variant.
OTOH, say selection favors a heterozygous population and things swing so that now 90% of the population is heterozygous for these genes and the remaining 10% is homozygous for A1. Now only 55% of the gene copies are A1 with the remaining 45% being A2.
In the second scenario genetic variability has increased because there is more even distribution between the two alleles. That is basically the point of what they are saying in the article.
It can never work. You cannot generate variety by killing. It's that simple.