This is really several related questions...
The equations of classical electrodynamics are time-reversal invariant. However, when we solve the equations for a particular system of charges it is common to take the solution with only retarded fields. The reason given for this choice is usually so that the solution agrees with our notion of causality, but this choice of boundary conditions is clearly not time-reversal invariant.
1) According to Feynman (Lectures on Physics, Volume II, chapter 20) taking the theory with time-reversal invariant boundary conditions 'in many circumstances does not lead to physically absurd conclusions'... but this implies than in some circumstances it does -- i.e. it should be possible to test via experiment that the choice of retarded fields only is the physically 'correct' one. What would such an experiment look like?
2) If such an experiment exists isn't it therefore wrong to say that the electromagnetic force is time-reversal invariant?
3) Is the time-reversal invariant theory Feynman mentions the same as Feynman-Wheeler absorber theory?
4) Is the same thing true in Quantum Electodymanics? That is, is there a condition equivalent to fixing the boundary conditions in the classical theory where only retarded solutions are considered in the calculation? And if so, does an alternative with time-reversal invariant boundary conditions exist? And if it does, does the distinction lead to experimentally testable differences?
Even if you can answer only one or a few of these questions it would be very much appreciated!
Update: I worry this question is already too long, but I'll try to be more specific about what I mean by time-reversal invariance. Start with some distribution of (classical) charges with their initial velocities and accelerations specified, and assumed to be isolated from the rest of the universe. The Feynman-Wheeler theory predicts their subsequent motion without any reference to fields. Then, at some time $T$ later measure all the new (positions, velocities, accelerations), reverse all velocities, and take this as a new initial condition for a second isolated system. Time-reversal invariance means that the dynamics of the second system should just be exactly the reverse of the first system (so that if I measure the state of the second system a time $T$ later and reverse all velocities again I get back to the initial condition of the first system).
I'm pretty sure this is true in the Feynman-Wheeler theory, but if I use the usual theory of electrodynamics and solve for both systems considering only retarded fields I guess it will be false. Therefore, in principle at least it is possible to tell the difference between the two theories experimentally; however, I'm not sure this is doable or has been done in practice.