After all these years, I’m still a noob when it comes to Rhino, but this one outright makes me feel stupid, I just don’t understand it:
With grid snap disabled, the second reference point is exactly where the cursor is, as I expect it. However, when I enable grid snap, then the second reference point snaps to in between grid points.
Why is that? 
Hi @feklee
Your scale operation starting point is the mid point of the line - which in this case falls exactly between two grid points. When you turn grid snap back on, it relates to the position of the scale starting point, not the . I’m not saying it’s logical, but that’s why. If you make a 6mm (or any even number) line, it will snap to the grid.
You can work around it by using SmartTrack - After setting your second reference point (the end of the line), move the mouse away from the end and then back, thus setting a smart track point at the end. That will force the grid snap to use the SmartTrack point as the reference instead of the first reference point. Try it out - it’s less complicated than it sounds in writing 
HTH, Jakob
Thanks for the explanation, but there is something missing!
What also works as a workaround, is pressing Shift while scaling. Weird, no?
To be honest, I think that behavior doesn’t make sense. I don’t comprehend the logic behind it. And I wonder if it was like that in the past.
I just checked with Rhino 7. _Scale1D behaves correctly there. So I assume, this is a bug.