There is no spooky interaction at a distance in relativistic local QFT. This is built in by construction, leading to the validity of the linked-cluster theorem. The entanglement of observables of far-distant parts of a quantum system does not enable causal faster-than light interactions (that misconception of some early Copenhagen-flavor interpretations was rightfully criticized by Einstein as "spooky action at a distance") but are long-range correlations between far-distant parts of a quantum system caused by its preparation by (usually some local) process (e.g., the preparation of a polarization-entangled photon pair by parametric down conversion using a BBO) and then waiting until the system has propagated long enough such that you can measure the photons' polarization state far away from their source and far away from each other. There's nothing acting spooky at a distance but just local interactions between the measurement device (polarization filter + photodetector) at each of the places. The observed 100% correlations of completely indetermined observables (i.e., completely unpolarized single photons) is there due to the preparation at the very beginning, and measuring the polarization of A's photon let's A immediately know what B will find (or maybe allready has found) when measuring his photon's polarization (with his polarizer directed in the same direction as A's), but she cannot transmit information to B by some faster-than-light mechanism. Even to check, whether the predicted 100% correlation really describes the findings of the experiment right, A and B have to exchange information about their measurement afterwards, and this they can't do with any faster-than-light signal. So, by construction of QED as a relativistic microcausal QFT there's no violation of causality by any "spooky action at a distance"!
