Inbreeding as an option to fixate aurochs traits?

Inbreeding or line breeding is common in animal breeding and basically is reproduction within a small population of closely related individuals reducing genetic diversity. It causes homozygosity and bears the risk of deleterious recessive alleles becoming fixed, reducing the biological fitness of the population and causing an inbreeding depression. But homozygosity also creates a uniform phenotype and therefore might be a method to create a more or less homogeneous, aurochs-like cattle population. In this post I am going to outline a little thought experiment I came up with about a week or so ago. 

The Heck cattle herd on Insel Wörth, Bavaria, is the only closed line within Heck cattle to date. The herd is experiencing inbreeding since the 1980s, since no cattle from other locations were included and consequently all herd members are closely related. Large, thick and well-curved horns are very common in the herd, and many individuals have a similar colour. But there are deviant examples as well, such as cows that are light-brown or reddish with almost no dark areas on head and neck, or with small and thin horns. There currently is one bull with rather steppe cattle-like horns on the island, and there was another cow with a rather straight horn curvature. Bulls with a greyish-beige saddle apparently appear regularly, I would say perhaps 50%, and I was told that the inheritance of horn sizes within the herd works by chance: very large-horned parents sometimes give birth to thin-horned individuals and vice versa. 
Why is the Wörth herd still not stable although it is a closed inbred line? 

Why do such small-horned, tan-coloured cows still appear in the Wörth lineage?

I thought about this and, as an amateur in genetics, came up with the idea that it might be back-crossing with heterozygous parents might be the key factor that made inbreeding inefficient in this herd. Back-crossing with a heterozygous parent makes little sense to me as there is a 0,5 probability that the parent passes on the undesired allele again, what, assuming that the offspring it was mated into is heterozygous as well, does not change anything, or, assuming that the offspring is homozygous, might even produce a heterozygous calf again and therefore make a step backwards. As a result, as far as I understand, a heterozygous individual cannot stabilize a herd, no matter how often it is mated to its offspring again. A homozygous bull, on the other hand, would of course. I think this might be true of "qualitative" traits that are controlled by few loci such as coat colour as much as "quantitative" traits that are controlled by many loci (and environment which we assume to be stable here) such as body size. 
True inbreeding lines created in the laboratory with rats, mice and other model organisms have a homozygosity level of roughly 99%. The important difference in this kind of inbreeding is that exclusively siblings are mated to each other. The rule of thumb is that at least 20 generations are needed to achieve this high level. The time span to do the same with cattle would be endless, but 99% is surely much beyond the extent of homozygosity that is needed for a optically more or less homogeneous aurochs-like breed. 

But it inspired me for thinking about sibling-matings as an option speed up the process of stabilizing the "breeding-back" results. I would start with a P-generation herd that is composed of bulls of one suited breed and cows of another breed that complement each other in terms of aurochs resemblance (or, alternatively, good-looking individuals of existing "breeding back" projects/breeds, wich would have a less predictable offspring on the other hand). The offspring they produce would form a new F1 herd with calf that have a more or less similar appearance to each other (except you work with very heterozygous parents like the already existing "breeding-back" cattle). They would be mated to each other and then a F2 herd would be created (by the way, the P and F1 would still exist and produce offspring, which would be consistently moved to the existing F1 and F2 herds etc.), and this generation would show a more or less continuous spectrum of the traits present in both breeds. Now, strong selection is ready to start. That scheme should be carried out until a herd is formed of which one can be sure that the individuals are sufficiently homogeneous, externally and genetically. I would then choose the best bull, fuse all the herds and remove all the other bulls. Using such a bull which would be mostly homozygous for the desired traits would then stabilize and improve the whole herd and create a larger, sufficiently homogeneous and stable line of aurochs-like cattle that can be used to improve other herds of any chosen breed/project. 
If this concept worked, there still would be some problems. First of all, one would need a lot of place for all those herds - like usual the larger the herds the better it is - and therefore also a lot of money. And of course there will be some animals that suffer from inbreeding depression and have to be selected out. This is called "purging", which is common in laboratory line breeding and also saved the highly inbred Chillingham cattle from extinction. The number of animals that would not be used for further breeding would be quite high, depending on how the strong the selection would be. They could be sold for meat production or to other breeders, as long as the looks of those cattle is appealing to them.
Surely the amount of inbreeding would be high for such a lineage, especially when using only two founder animals as suggested above (personally, I would use a good Wörth-Heck cow with large horns and a very large, slender and well-shaped bull, perhaps one of these "Maltese Oxen").  Probably most people would fear that selecting only against deleterious alleles and only for aurochs-like looks would reduce the hardiness or resistance to diseases. I don't really worry about the hardiness of the cattle, as all aurochs-like cattle seem to be very hardy. However, the resistance against diseases might get reduced due to genetic drift if the other breed is not as resistant against one particular disease than the other. But as described above, my purpose is not to create only one closed line that looks as aurochs-like as possible, but to use animals of such a very stable line to increase the similarity with the aurochs of other aurochs projects, or simply breeds of interest, as a whole. Apart from that, there could be other, parallele projects doing the same but using other breeds, and uniting those stable lines would increase the gene pool. Using homozygous breeding bulls (but also cows of course), will speed up the fixation of the desirable features for any project or breeder who wants to have very aurochs-like cattle without having deviant traits appearing regularly. 

I think the traditional scheme of just crossing and selecting the best-looking animals and selecting out the bad-looking ones is not very effective or at least takes a lot of time. Especially quantitative traits such as horn size or body size show only a slow response to selection (see here, for example).  Whether or not such a concept is practically feasible or effective, or if any of the existing projects/breeders would be interested to try it, it is a nice little thought experiment of mine as being a total amateur in genetics. If there is anything incorrect or imprecise in this text, or if the idea is flawed by any technical reasons, I would be pleased if somebody more knowledgable would point it out to me.