Look the picture which gives the theme of this thread here, what you might call OP pic related. You can see the horizontal axis is time, and now I'm telling you that the curve is a gravitational potential E(t). Specially, the time is the proper time of a clock in the universe. Let the left and right sides of plot be past and future timelike infinity respectively. Place some point mass at future timelike infinity. Usually the Newtonian gravitational potential goes as 1/r, but here we have E(t) so the potential goes as 1/t. From freshman-level kinematics, it is easy to see that if we take any two surfaces of constant proper time, call them t_1 and t_2, these two surfaces will accelerate away from each other.

For instance, with E(r) as the familiar example, if you release a bowling ball out by Uranus and another by Venus, observers at the locations of each bowling ball will see the other start to accelerate away from their own inertial frame by their bowling ball, because the ball by Venus is much deeper in the curved well.

Since the images of supernovae which appear in our telescopes (from the existence of dark energy is inferred) are located on surfaces of constant proper time much earlier than the hypersurface of telescopes in the present, those supernovae images will be observed to recede. All you have to do is ask, "What curve parameter is required to fit the dark energy data?" Then you simply hypothesize the mass whose gravity well will fit that parameter when located at future timelike infinity, and it will perfectly describe the DE spectrum without having to assume DE to begin with. The only thing difficult is to come up with a way to measure distance from future infinity, but I have already come up with it: use infinity-hat to measure the distance. It makes perfect sense. Even though it makes perfect sense in analysis, physics requires no rigor whatsoever, and it makes even more sense in cosmology than the perfect sense it already made in analysis.