One needs three non linear points to get a curve with SmoothLine & BEZIER_INTERPOLATION, does one not?
Bruno has only the two end points.
He could calculate a point mid-way point on the straight a-b line, and then offset it a certain amount (progressively or randomly).
That would give three points for SmoothLine.
It would also be a lot of code, and may be a bit crude.
If the lines are very long, one could calculate the lines in a coordinate space where they are straight, densify them, and then project them to the target space where they would be curved.
For example, plot lines A-B, A-C, A-D, A-x in an azimutahal space centered on A.
Densify the lines to 'pin down' the path geographically.
project to a target space (say, Mercator) for display
the lines should be curved....
As I said, the lines would have to be long enough for the earth's curvature to matter.