Motivated by predictions made using a bond valence sum difference map (BVS-DM) analysis, the novel Li-ion conductor Li2Mg2P3O9N was synthesized by ion exchange from a Na2Mg2P3O9N precursor. Impedance spectroscopy measurements indicate that Li2Mg2P3O9N has a room temperature Li-ion conductivity of about 10(-6) S/cm (comparable to UPON), which is 6 orders of magnitude higher than the extrapolated Na-ion conductivity of Na2Mg2P3O9N at this temperature. The structure of Li2Mg2P3O9N was determined from ex situ synchrotron and time-of-flight neutron diffraction data to retain the P2(1)3 space group, though with a cubic lattice parameter of a = 9.11176(8) A that is significantly smaller than the a = 9.2439(1) angstrom of Na2Mg2P3O9N. The two Li-ion sites are found to be very substantially displaced (similar to 0.5 angstrom) relative to the analogous Na sites in the precursor phase. The non-molten salt ion exchange method used to prepare Li2Mg2P3O9N produces a minimal background in powder diffraction experiments, and was therefore exploited for the first time to follow a Li+/Na+ ion exchange reaction using in situ powder neutron diffraction. Lattice parameter changes during ion exchange suggest that the reaction proceeds through a Na2-xLixMg2P3O9N solid solution (stage 1) followed by a two-phase reaction (stage 2) to form Li2Mg2P3O9N. However, full Rietveld refinements of the in situ neutron diffraction data indicate that the actual transformation mechanism is more complex and instead involves two thermodynamically distinct solid solutions in which the Li exclusively occupies the Li1 site at low Li contents (stage 1a) and then migrates to the Li3 site at higher Li contents (stage 1b), a crossover driven by the different signs of the local volume change at these sites. In addition to highlighting the importance of obtaining full structural data in situ throughout the ion exchange process, these results provide insights into the general question of what constitutes a thermodynamic phase.