No real design or engineering was involved beyond having an idea of what we wanted and trying to select for that trait in breeding and offspring. We've just rolled the genetic dice repeatedly and selected the outcomes we preferred.
> we absolutely should continue doing what we're already doing.
This is very unlike what we have been doing, and it's hardly been an unqualified success - lots of traits we have selected for (eg what defines a particular dog breed) have come with real trade-offs (at least for the animals: arthritis and fits, respiratory conditions, uncontrolled drooling, eye infections, ...)
And? That's just a particular choice of tool. I do the same thing when I'm working on a particularly tough problem; I might try five or ten different approaches before finding one with the right combination of advantages and trade-offs.
> lots of traits we have selected for [...] have come with real trade-offs
That's engineering.
It wasn't a choice, it was literally the only method we had available.
> I do the same thing when I'm working on a particularly tough problem; I might try five or ten different approaches before finding one with the right combination of advantages and trade-offs.
But I bet you don't do it by a process of natural selection, setting criteria and evolving solutions through an at-best semi-guided generally-black-box process that you don't really understand, and where you're only able to understand a portion of the full complexity of the results.
> That's engineering.
No, that's outcomes. Engineering would be understanding the tradeoffs in advance and choosing them with consideration of the outcomes. Not only do we not understand the tradeoffs in advance, we're also not able select them reliably. If we have that ability now via breeding (not that I'm aware of, but I'm no expert on breeding technology) then we've only had it for a very short time relative to how long we've been doing this.
You're not giving enough credit to the people that designed domestic animal breeds. Modern working dog breeds were absolutely engineered for particular tasks; target characteristics were selected up-front, size and hair and temperament and so on, and dogs and breeds were chosen to start from and bred in predictable ways to increase the likelihood of obtaining the desired characteristics. Animal husbandry is only "random" in the same way that sports is: There's absolutely some variation in outcome in each individual game, but if you put a professional NFL team against even a college team there's not much doubt about the outcome of a best of five. Or it's random in the same way a loaded die is: individual trials may be random, but if you consistently use it for the important rolls you'll clean up.
Also, I'd just like to say that software engineers would be ecstatic if they had a reliable way to accurately predict the tradeoffs involved in many architectural choices.