As I see it, comparing for equality would be the typical usage for a reference waveform (with a live second channel, this is likely to be different). Being able to size and position the reference in both the time and voltage dimensions would add flexibility. I have no reservations with respect to this argument. This does come at the added expense of complexity however in that we need additional controls for T/Div (if you really want to change the time domain), V/Div, Trigger Pos and Vert Pos related to reference. We would also need to consider cursors (V1, V2, T1, T2 and Trigger Pos) as these can apply to only one set of domain parameters. Still I think there may be ways to solve all of these issues if there is consensus that this adds real value to real problems.
More important perhaps is that we should not change the default behavior in such a way that typical usage is obscured and this is where I struggle to see where you want to go with “retaining XML parameters”.
Don’t think for a second however that I do not appreciate the ideas however challenging they may be.