I don't know that it really matters, but I have to ask. Are you asking about human or other populations? Dogs, which tend to be inbred, may show different patterns, for example than "wild" fruit flies. Many crop species are highly inbred.

Well I'm asking for human but this is an interesting point.

Perhaps it would be helpful to define what a heterozygous or homozygous CNV actually is. A copy number variant can actually arise in several ways. For example, not all sequence objects that give rise to a CNV are in cis, so het versus hom might not actually make sense for many CNVs.

Right. And partial overlap of "independent" CNVs could lead to a number of switches from het to hom across the CNV region.

From personal experience with human Affy microarray data, mostly hemizygous states are observed. But, as noted above, this has almost nothing to do with underlying biological mechanisms. For example, a conservative rearrangement to another chromosome could produce something that you would consider a homozygous deletion functionally, but you would not see that based on SNP array data.