I really wish an approach like clojure's with core.logic was more common. It's been fairly often in my career that I run across small semi-contained problems within a larger codebase that would be very well suited to prolog, but nowhere near worth using a separate language and toolchain for.
Having that in the standard library of more languages would be so useful. I know minikanren exists, but the page kind of touches on this: it has usually been implemented as an exercise. Minikanren libs in languages where I've wanted them have been incompletely documented, and lacking some of the practical additions.
I agree that core.logic seems to have found a sweet spot as a pragmatic embedded domain-specific language. If core.logic had tried to replace more of the Clojure functionality, or had tried to be a stand-alone Everything Language, I doubt it was have been attractive to Clojurers.
I also agree that having something similar to core.logic in many other languages would be useful.
Most miniKanren implementations in most languages were created in order for the implementor to better understand miniKanren or logic programming, or because the implementor wanted to experiment with miniKanren-related reseach ideas.
This split between a pragmatic implementation versus a small, simple to implement/understand/teach/hack implementation may be another legacy of miniKanren coming from the Scheme community, which has struggled with this dichotomy for decades.
Having that in the standard library of more languages would be so useful. I know minikanren exists, but the page kind of touches on this: it has usually been implemented as an exercise. Minikanren libs in languages where I've wanted them have been incompletely documented, and lacking some of the practical additions.