this doesn't seem like a bug in nsd. rather, it exposes a problem in
yeti mzsk. while it's true that nsd should fall back to axfr in this
case, that "bug" is almost unrelated to the prime case which is mzsk
itself. this warrants further discussion.