Australian Dingos maybe used to be (semi-)domesticated before they were introduced to Australia, originally having been bred in China, then gone 'wild' between 5000 and 18,000 years ago (https://en.wikipedia.org/wiki/Dingo#Origin_and_genetic_statu...). There is some research into the genome on that page, but AFAIK there is no whole genome assembly yet. You can tame those but wild dingoes are very dangerous animals, as many Australian tourists can tell you.

Zebras, on the other hand, are generally thought to be very hard to domesticate, with only a few successes of taming in the last hundred years, and zero success for domestication: https://en.wikipedia.org/wiki/Zebra#Domestication

This could have relevance to dog breeds originally selected for fighting or baiting, then subsequently selected for companionship. Maybe if a fighting or baiting dog is n generations into social breeding it can be considered safe?

Dogs are kind of a weird case, from an animal population standpoint. On one hand, there are a ton of 'purebred' populations that undergo very specific selection and which are fairly genetically isolated from the rest of the species. These are your fighting dogs, your hunting dogs, your ratters, your herders, your companions, etc etc. On the other, there's this huge pool of castoffs and mutts, which aren't really undergoing any concerted deliberate selection at all. However, the castoffs from the specific breed dogs are constantly being shunted into the general population.

So to answer your question in that context: Yes, it would be more than possible to take a population of dangerous fighting dogs and use them as the base stock for a pet breed, provided that the breeding program was well designed. However, the general mutt population will always have fighting dog traits mixed into it so long as there are fighting breeds around.