Ok, so as I guessed, multiQC fails with the tiny test data, but not mega tests. So we have 2 solutions. Either we ship with 1.33 but conda fails, or we disable multiQC in the failing tests and we ship with 1.34