If we define the ortho-camera's width to be the major dimension (Sensor Fit = Horizontal), then the width of the ortho-camera's viewport is set to 1 blender unit (in camera-space). Otherwise the height's viewport is set to 1 blender unit (in camera-space).
The orthographic scale now results from the ratio of the above defined blender unit (in camera space) to 1 blender unit (in global space).
By using elemental trigonometric maths a unit square in global space would exactly fit in the camera's viewport (width), if the value of the orthographic scale is exactly √2 = 1.41... .
So if you double this value to 2*√2 then 2 unit squares will exactly fit in the camera's viewport width (while the viewports camera width still remains 1 blender unit in camera space).
With the knowledge, that the ortho camera's width is always 1 blender unit, we can only vary the camera's visible viewport content by
- changing the orthographic scale
- and/or changing the resolution of the camera's viewport width and height
With those two main parameters it is possible to adjust the ortho camera exactly to your specific needs.