Guillardia theta assembly file is around 50mb. This wasn't a problem when the tests were downloading it every time from NCBI. However, now we are mocking NCBI api with a cached version of the file (so that tests don't break every week due to NCBI side changes). So this file is now stored in git-lfs, and with TravisCI testing we quickly reach the 1GB/month free GitHub git-lfs bandwidth. Maybe there's a small virus, bacteria or DNA region on NCBI with a much smaller assembly.