This is the claim Tommasini makes in Reality, Measurement and Locality in Quantum Field Theory:"Two spin $1/2$ particles, A and B, are created in coincidence in a spin-singlet state, and are detected by the detectors $O_A$ and $O_B$ in opposite directions... The EPR argument, as described above, and (as far as I know) all the subsequent treatments of the EPR paradox, have assumed that it was actually possible to prepare a system of two entangled particles. However, I have recently proved that this assumption is not correct... In fact, the Standard Model of Particle Physics predicts that it is not possible to produce a state having a definite particle content: given the process that produces A and B alone, QFT theory predicts a nonvanishing and finite probability for the creation of A and B plus additional photons".
She goes on to say that "the EPR+Bell proof of nonlocality is removed" because "for the EPR argument it is crucial that the measurement on A implies a certain prediction for B without disturbing B". But spurious photons potentially produced along with A and B make any prediction uncertain.
Is it true that in QFT one can not prepare states with prescribed number of particles? Does it follow that above analysis of EPR is correct? QFT is manifestly relativistic, so it makes sense that quantum non-locality is "removed", and Tommasini reproduces the usual QM correlations for EPR using a Feynman integral QFT calculation, so it seems consistent. But this diverges sharply from the usual explanation of EPR.
EDIT: In a companion paper there are some details on computations and agreement with experiments:"the case of the EPR experiments that have been performed up to now the QED prediction for the correlations is very close to that obtained in Quantum Mechanics by ignoring the soft photons, so that it can still agree with the data within the experimental errors. However, even a very small probability for soft photons creation is sufficient to forbid any certain prediction for the measurement on B as a consequence of the measurement on A".
Apparently, soft photons do exist (her source is Weinberg's text), and they do affect QED predictions:"Even though soft photons are not detected, the possibility of their emission must be taken into account in the calculation of the scattering amplitude". Entanglement and the infrared structure of QED discusses QED violations of Bell inequalities:"We might consider that they started with the studies of the effect of the QED spin-spin interactions on the entanglement and the violation of Bell Inequalities due to QED... The objective of this work is... to characterize the effect of soft photons on the entanglement of two charged qubits..."
So I guess the answer to the first question is affirmative. I am still not clear though why small QED corrections to QM correlations entirely "remove" non-locality.