Having complete tests for this repository would be crazy (because some of the stuff we run could take weeks). But we should create a small-ish runcard per runmode so that we can check that things can run.
Even better if we do it regression-test style, so that we ensure that when results change we know why it is.