I'm saying that these sorts of things (like with the sexual preference feature) are likely to be universally applied, and can be used as a platform from which behaviors like nesting can be applied, rather than being some sort of hard-coded one-off tag, the way that cats, for the longest time, had their vermin-hunting and their "dwarf adopting" behaviors married to one another.
Expanding this into a more robust set of actions can allow for a wide variety of interesting (and realistic) animal behaviors, but also the capacity for modding interesting societal behaviors. (Cheetahmen migrants racing on the surface above the fortress when Spring is in the air...)
The problem of crocodile training lies more with the fact that domestication has essentially always occurred with instinctively social creatures which humans could exploit through breeding animals that saw humans as "members of their herd/pack/whatever". A wolf is social and intelligent enough to learn new behaviors from pack members. A crocodile is a solitary hunter that relies purely upon instinctively bred behaviors and is not evolved to learn behaviors from parents that don't care for their young, regardless. That sort of problem lies outside the scope of this suggestion.
To get back more to what I was previously talking about, this has uses with fully wild animals, as well. The most obvious is that wild animals currently don't lay eggs at all because they don't make nests, but wild animals performing courtship behaviors in the wild would be interesting as pure simulation. Consider if, instead of fish just sitting in a river and occasionally having more fish fry appear near a female, your rivers would have salmon start attempting to jump up rapids before laying eggs and then flipping over and dying. It's one of those verisimilitude things that DF thrives upon.