Currently, we cannot aggressively parallelise integration tests due to possibility of cleanup routines for each test instance stepping over each other.
3 things are probably needed:
[ ] We need to pass a unique key id for the run to every workflow execution and target the cleanup on that.
[ ] Set up a backup cleanup action that would ensure cleanup of objects outside of the test teardown routine (e.g. if tests error out).
[ ] Set up a final backstop running every week or so to cleanup everything else that might get stuck accidentally so we don't overrun our CI bill.
Currently, we cannot aggressively parallelise integration tests due to possibility of cleanup routines for each test instance stepping over each other.
3 things are probably needed: