Right now we use a parent test runner and reusable child workflows for defining our safe PR mechanism. This has some downsides, notably that all our tests are in the same suite and it confuses various monitoring. Instead, we could define an action to do a lot of the work so that splitting up our monolithic runner is more feasible