We have a large number of tempest dsvm job configurations and has many options. For most plugins/drivers/etc. it is reasonable to rely on third-party ci to make sure things don't break. But there are lots of of other things like keystone v2/v3, admin/non-admin, slower tests, that are best handled by post-merge or periodic runs. This has been discussed before but has two problems: we have no way of tracking or visualizing these jobs, and it is not clear who should pay attention to them. We need to solve these issues and move away from the current situation.