There is considerable evidence for local adaptation in nature, yet important questions remain regarding its genetic basis. How many loci are involved? What are their effect sizes? What is the relative importance of conditional neutrality vs. genetic trade-offs?  Here we address these questions in the self-pollinating, annual plant Arabidopsis thaliana. We used 400 recombinant inbred lines (RIL) derived from two locally adapted populations in Italy and Sweden, grew the RILs and parents at the parental locations, and mapped quantitative trait loci (QTL) for mean fitness (fruits/seedling planted). We previously published results from the first three years of the study, and here add five additional years, providing a unique opportunity to assess how temporal variation in selection might affect QTL detection and classification. We found ten adaptive and one maladaptive QTL in Italy, and six adaptive and four maladaptive QTL in Sweden. The discovery of maladaptive QTL at both sites suggests that even locally adapted populations are not always at their genotypic optimum. Mean effect sizes for adaptive QTL, 0.97 and 0.55 fruits in Italy and Sweden respectively, were large relative to the mean fitness of the RILs (approximately 8 fruits/seedling planted at both sites). Both genetic trade-offs (four cases) and conditional neutrality (seven cases) contribute to local adaptation in this system. The eight-year dataset provided greater power to detect QTL and to estimate their locations compared to our previous three-year study, identifying one new genetic trade-off and reclassifying one QTL from genetic trade-off to conditionally neutral.