Photons have no identities (they are exchangeable), the fact that you need to mark them (to differentiate a random one from an entangled one) is a consequence o

Photons have no identities (they are exchangeable), the fact that you need to mark them (to differentiate a random one from an entangled one) is a consequence of your choice of model (QM) that you need to keep track for the computation. Field theories are not subject to such constraints. You can interfere a photon with its entangled photon. You can interfere it with itself too. As soon as your first photon get absorbed, in your calculations about the second photon you incorporate the new information you got about the first, and you update your probability distribution accordingly (but that's not a physical process it's just your way of computing the probability). When the first photon is absorbed the second is not necessarily so and he can go its merry way. (If you use the information you gain from the observation of the first photon about its polarization, and you have set-up your experiment so that you have time to orient a polarizer in the right orientation so that your photon can pass, you can catch the second photon 100% of the time). There have been tons of experiments, where you entangle the photons n-way, split them multiple times, and interfere them in various ways. Feynman propagator conserve various quantities allows you to evolve your probability distributions forward and backward in time. When you compute the conditional probabilities that correspond to your observations, you can shift them all back to the same instant where the entangled pair was produced. You can't expect to change conserved quantities, because they are continuously conserved. For quantities that are not subject to specific conservation law like for example the time of detection of the photon, you can compute correlations, but there are also subject to conservation of information. You could also theoretically imagine experiments where you entangle photons and have them interact with entangled pairs of electrons. But it's more for the intellectual pleasure of trying to do complex calculations correctly. Personal unorthodox conviction : It's not the way the universe does the computation. The universe has its (finite) configuration it is in which is continuously (and locally) evolved through time, but which as an observer inside the universe you can't have access to (think of it like the seed of a random number generator). It doesn't need the results of your experiments to evolve through time. The way you can compute the same way the universe does is through Monte-Carlo simulations of the joint probability, but from inside the universe you have to generate a lot of (possibly infinite) candidate trajectories (aka possible universes) and do the sum (or filter to match the observation to the universe you are in).