Genetic diversity cannot decline when you start from a single pair. With sexual species, a single pair is as low in diversity as you can get without going extinct.
I don't see why diversity couldn't decline. A population of two could have up to four alleles for any given gene, so if any of these were lost the population would have less diversity. This wouldn't drive them extinct. It might leave them less able to cope with environmental change, but you could wind up with a population that's totally homozygous for one gene or other that's still producing viable indivduals.
And, if I understood the article correctly, this happened less than expected by theoretical modelling in this population. In such a tiny population all it would take is for a couple of homozygous individuals to be very successful and one allele would become by far the dominant one in the population. Instead, they found the population had maintained many of the different alleles, and they were present in a variety of combinations.
What this means for Noah's ark is that single pairs could give rise to viable populations (the individuals should probably be as distantly related as possible), but genetic measures of diversity would still reveal that there were only 4 alleles max per gene. This is not what we see today.
Four alleles plus any additions through mutation in the last few thousand years. This isn't meant to be supporting evidence for the idea of the Ark, though. Jar had just asked for examples of pregnant individuals giving rise to viable populations. This was the closest I know of.
™ Version 4.2