Background: The application of target capture with next-generation sequencing now enables phylogenomic analyses of rapidly radiating clades of species. But such analyses are complicated by extensive incomplete lineage sorting, demanding the use of methods that consider this process explicitly, such as the multi-species coalescent (MSC) model. However, the MSC makes strong assumptions about divergence history and population structure, and when using the full Bayesian implementation, current computational limits mean that relatively few loci and samples can be analysed for even modest sized radiations. We explore these issues through analyses of an extensive (>1000 loci) dataset for the Australian rainbow skinks. This group consists of 3 genera and 39 described species, which likely diversified rapidly in Australia during the mid-late Miocene to occupy rainforest, woodland, and rocky habitats with corresponding diversity of morphology and breeding colouration. Previous phylogenetic analyses of this group have revealed short inter-nodes and high discordance among loci, limiting the resolution possible. A further complication is that many species have deep phylogeographic structure - this poses the question of how to sample individuals within species for analyses using the MSC. Results: Phylogenies obtained using concatenation and summary coalescent species tree approaches to the full dataset are well resolved with generally consistent topology, including for previously intractable relationships near the base of the clade. As expected, branch lengths at the tips are substantially overestimated using concatenation. Comparisons of different strategies for sampling haplotypes for full Bayesian MSC analyses (for one clade and using smaller sets of loci) revealed, unexpectedly, that combining haplotypes across divergent phylogeographic lineages yielded consistent species trees. Conclusions: This study of more than 1000 loci provides a strongly-supported estimate of the phylogeny of the Australian rainbow skinks, which will inform future research on the evolution and taxonomy of this group. Our analyses suggest that species tree estimation with the MSC can be quite robust to violation of the assumption that the individuals representing a taxon are sampled from a panmictic population.