Re the relative orientation of two sets of atoms: hjkl might be a bit coarse,* because naturally it runs in increments. If this were to coarse (additional input by other users welcome), perhaps a) this could be complemented by dials of the three axes (similar to e.g., inkscape "rotate this selected shape by x degrees in in clock/anticlockwise direction), or b) right away with an input for the Eulerian rotation matrix. I think if not a), version b) could be “too much effort” for Avogadro2.

* Of course not to be called vim tool. There are times vim is more suitable, and other times Emacs (e.g., orgmode) for different purposes.