Tempest reached an important milestone recently when the whole suite became a gating job for all projects. This is good, but as
1. more tests pour in to projects 2. more projects have multiple API versions 3. more projects become Integrated 4. Tempest moves towards a real acceptance test for OpenStack
it is unlikely that "gate all projects on the full test suite of every other project" is sustainable. We should be able to come up with a better strategy to decide which parts of Tempest should be gating and make gating test coverage more modular while minimizing the risk of regression.
