Many genes harbor a loss of function variant with no clear negative effect. If you look at a population, the frequency at which genes carry any predicted deleterious variant is non-negligible. This can be used to estimate how extreme it is for a monogenic disease study to observe a gene with multiple deleterious variants in it.
However, most genes work in gene pathways. It may be that a single gene being disrupted does not disrupt the pathway substantially, but multiple genes being disrupted in the same pathway can have large effect.
The null model would be there there is no correlation between being a deleterious variant carrier between two genes. However, if you look at cases and controls and at gene pathways, it may be in cases have a positive correlation while controls have a negative correlation.