1) Does the ability to breed between individual members of A and B decline as generations pass and the genetic differences grow (expressed as an "impregnation probability")?
As a breeder of dogs and parrots, I have some expertise.
As you surely must know, the definition of a species has nothing to do with the ability to interbreed. Hybridization between species in the same genus (and occasionally even beyond that) is routine in captivity and common enough in the wild to be unremarkable. The aviculturist market is rife with hybrid macaws, Amazons, cockatoos, conures and parakeets. Hybrid felines like the ocicat are now in vogue, and mules and zebrasses have been around for years. (Not all of these hybrid offspring, such as the mule, can produce a subsequent generation, but most can and do.)
Hybrid conures and macaws have been spotted in the wild in South America, and hybrid
Pheucticus grosbeaks are obliterating the distinction between the black-headed and rose-breasted species in North America.
What is more usually the factor that separates populations is the evolution of incompatible courtship rituals. Many birds rely on displays of color so a closely-related species does not give off the right signals. Ornithologists report that the hybrid conures in the rain forest have settled into distinct communities that only breed among themselves--although macaws appear to be avian hippies with a certain enthusiasm for inter-species dating.
Lions and tigers can be successfully crossbred using AI, but never from copulation. Tigers need to be clawed during courtship in order to trigger the proper response... and lions can't stand it.
2) If #1 is true, is this effect seen in dog breeding?
Absolutely not. A female dog in estrus releases pheromones that attract every male dog within a wide circle, and she will compulsively mate with all of them so that each one will assume that the puppies are his and therefore protect them. Dogs of vastly different sizes have been known to mate and produce offspring. A few years ago a lab popped a litter of half-chihuahua puppies without anyone even realizing she was pregnant.
Dogs and wolves (as I'm sure you know) are two subspecies of
Canis lupus, and under the right circumstances they crossbreed in the wild, and routinely in captivity. Both animals hybridize with
Canis latrans, the coyote. The eastern US is currently seeing a migration of gigantic wolf-coyote hybrids from Canada, just in time to help with our deer problem.
In other words, are there disparate human races which have a more difficult time forming a zygote together due to their genetic differences?
Ignoring the past for a moment, this is emphatically not true today. Since the Industrial Revolution made long-distance travel commonplace, whatever neatly defined gene pools might have once existed have been obliterated. We can still identify the location of a person's ancestors from his DNA of course, and sometimes even from physical appearance, but we can also see a sizeable percentage of that DNA from outside sources. To use the language of us dog breeders, no humans are purebreds; we are all mongrels.
As for the past,
Homo sapiens first migrated out of Africa around 50KYA. Sometime between then and now, their migration routes resulted in four relatively isolated populations in sub-Saharan Africa; Australia; eastern Asia and the Americas; and western Asia, North Africa and Europe. Nonetheless as they came in contact with each other there were certainly no impediments to interfertility.