Insecticide resistance provides both a pressing threat to the control of vector-borne diseases and insights into the remarkable capacity of natural populations to show rapid evolutionary responses. Malaria control remains heavily dependent on deployment of insecticides, primarily in long lasting insecticide treated nets (LLINs), but resistance in the major malaria vectors has increased over the last 15 years. Identifying genetic mechanisms causing high-level resistance in mosquitoes, which may almost entirely overcome pyrethroid efficacy, is crucial for the development and deployment of potentially resistance-breaking tools. Using the Anopheles gambiae 1000 genomes data we identified a very recent selective sweep in Ugandan mosquitoes which localized to a cluster of cytochrome P450 genes. Further interrogation revealed a haplotype involving a trio of mutations, a point mutation in Cyp6p4, an insertion of a partial Zanzibar transposable element (TE) and a duplication of the Cyp6aa1 gene. The mutations appear to have originated recently in An. gambiae from the Kenya-Uganda border region, with stepwise replacement of the double-mutant (Zanzibar TE and Cyp6p4-236M) with the triple-mutant haplotype (including Cyp6aa1 duplication), which has spread into the Democratic Republic of Congo and Tanzania. The triple-mutant haplotype is strongly associated with increased expression of genes able to metabolise pyrethroids; is strongly predictive of resistance to pyrethroids but importantly, appears less effective against LLINs co-treated with the synergist piperonyl butoxide (PBO). Frequencies of the triple-mutant haplotype remain spatially variable even within countries, suggesting an effective marker system to guide deployment decisions for limited supplies of PBO-pyrethroid co-treated LLINs across African countries.