The ability of insect pests to rapidly and repeatedly adapt to insecticides has long challenged entomologists and evolutionary biologists. Since Crow's seminal paper on insecticide resistance in 1957, new data and insights continue to emerge that are relevant to the old questions about how insecticide resistance evolves: such as whether it is predominantly mono- or polygenic, and evolving from standing vs. de novo genetic variation. Many studies support the monogenic hypothesis, and current management recommendations assume single- or two-locus models. But inferences could be improved by integrating data from a broader sample of pest populations and genomes. Here we generate evidence relevant to these questions by applying a landscape genomics framework to the study of insecticide resistance in a major agricultural pest, Colorado potato beetle, Leptinotarsa decemlineata (Say). Genome-environment association tests using genomic variation from 16 populations spanning gradients of landscape variables associated with insecticide exposure over time revealed 42 strong candidate insecticide resistance genes, with potentially overlapping roles in multiple resistance mechanisms. Measurements of resistance to a widely-used insecticide, imidacloprid, among 47 L. decemlineata populations revealed heterogeneity at a small (2 km) scale and no spatial signature of origin or spread throughout the landscape. Analysis of nucleotide diversity suggested candidate resistance loci have undergone varying degrees of selective sweeps, often maintaining similar levels of nucleotide diversity to neutral loci. This study suggests that many genes are involved in insecticide resistance in L. decemlineata and that resistance likely evolves from both de novo and standing genetic variation.