We report new calculations of interstellar 15N fractionation. Previously, we have shown that large enhancements of 15N/14N can occur in cold, dense gas where CO is frozen out, but that the existence of an NH + N channel in the dissociative recombination of N2H+ severely curtails the fractionation. In the light of recent experimental evidence that this channel is in fact negligible, we have reassessed the 15N chemistry in dense cloud cores. We consider the effects of temperatures below 10 K, and of the presence of large amounts of atomic nitrogen. We also show how the temporal evolution of gas-phase isotope ratios is preserved as spatial heterogeneity in ammonia ice mantles, as monolayers deposited at different times have different isotopic compositions. We demonstrate that the upper layers of this ice may have 15N/14N ratios an order of magnitude larger than the underlying elemental value. Converting our ratios to delta-values, we obtain delta(15N) > 3,000 per mil in the uppermost layer, with values as high as 10,000 per mil in some models. We suggest that this material is the precursor to the 15N `hotspots recently discovered in meteorites and IDPs