In order to solve DRIVERS-2715, we ensured that clusters were deleted at the end of each run. However, this led to projects remaining open, and our project limit being exceeded.

In DRIVERS-2734, we also ensured that projects were deleted at the end of each run by handling the CANNOT_CLOSE_GROUP_ACTIVE_ATLAS_CLUSTERS status and waiting until the cluster is deleted. However, this resulted in clusters failing to start before the one hour limit, and also led to projects being left open.

I believe what we need to do is reap all of the projects at once as a separate job, that depends on all the other jobs:

astrolabevenv/${PYTHON_BIN_DIR}/astrolabe projects \
            delete-all --org-id=${atlas_organization_id}