Implementation looks good. But can you add the equation to the docstring. That makes it easier to read without having to understand code (and Scipp idiosyncrasies like dtype conversions).

What do you mean by 'aligned' here? I understand that word to mean that the beam is in the sample plane, i.e., the incident angle is 0. Is sample_rotation defined as the incident angle? If so, just use that here!

What do you mean by 'aligned' here? I understand that word to mean that the beam is in the sample plane, i.e., the incident angle is 0

Yes you're right, this is unclear.

In this context, what I mean by "aligned with the beam" is exactly that rotation_angle equals the angle of incidence on the sample (which makes the statement above a bit circular).

But I thought it is good for the user to be aware that we make this assumption.

Maybe it would be good to introduce a quantity called angle_of_incidence, to clearly distinguish that from sample_rotation.

Could do.
Personally, I'm quite partial to using images to show these quantities. Basically, a modification of the images in https://scipp.github.io/scippneutron/user-guide/coordinate-transformations.html would be useful here. Maybe we should add them somewhere centrally in the ESSreflectometry docs. Especially since some definitions differ from ScippNeutron.