In particle physics we use Feynman diagrams to calculate the probabilities of interaction of two partilces. This is a Feynman diagram of electron positron annihilation , dominant for energies low enough so that more particles cannot appear. At high energies a lot more particles come out than two gammas.
Now the probabilities of interaction come from the prescription for Feynman diagrams and two photons are indistinguishable in this calculation. They do not have "memory" of the initial two particles other than the angular distribution, a probability plot, computable from the diagram. The diagram can be read and is valid for calculations in reverse time.
I cannot understand what you mean by reversible. If you mean can two photons scatter off each other and produce an electron positron pair, yes, one could compute the probability amplitude for that. Usually this happens often in detectors with the second gamma a virtual one from the field of an atom or a nucleus.
If you mean the two original gammas of e+e- annihilation, they will never meet again as they depart with the velocity of light, and the probability of meeting another one that has come from an e+e- annihilation is very very small unless it is two gamma beams designed to see this ( but why would one want to do that) . They are talking of gamma gamma colliders, but at much larger energies to generate a lot of particles.