Failure testing

Failure scenarios need to be tested. Ideally, this happens automatically on a semi-regular basis to make sure that there is a record of what might happen to the mesh in a specific scenario.

Each of these tests should be run with and without a HA configuration.

- Node failure with the control plane on it - destination, identity and proxy-injector are the most important here.
- Node failure with the data plane on it - does traffic continue to be routed to this pod?
- Complete cluster failure of all nodes
- Loss of communication to the api-server

For each test, please document the failure scenario, what degradation occurred and the impact on the users.

Note: for the purpose of this task, node failure means that the node was not shut down cleanly and the pods scheduled on that node are no longer running.

Start by locating the repository's existing Kubernetes failure or integration-test entry points and its HA configuration setup. Exercise each listed node, cluster, and API-server failure with and without HA, then document the scenario, observed degradation, and user impact for every test.

Written by the indexing model from the issue text.