I just stared at a globe, and it’s possible, but not from a geostationary orbit.

Think about a hemisphere projected onto a 2d circle - you can see all points on the surface of the hemisphere on the projection.

Logically it must be possible to select a hemisphere, such that any two random points on a sphere lie on that hemisphere, and would thus be visible in the projection.

The only edge cases is two points that are diametrically opposite - i.e north and south pole, which would only both be visible at infinite distance.

Points nearly diametrically opposed would only be visible from extreme distance.

This doesn’t take into account inconvenient mountains, but they can easily be moved.