This PR addresses 3 outstanding issues with the current broken tests:
Pulumi Yaml Program test is flakey as specs within it interact with ConfigMaps of the same name. We can run this sequentially, but I've fixed this by dynamically generating the require Pulumi Program CRs and randomising ConfigMap names.
The stale state test is flakey and needs to be run sequentially by decorating it with Gingko's Ordered decorator.
Usage of pulumi-kubernetes is pinned to v3.26.0 of the provider. v3.27.0 was recently released and caused our previously passing tests to fail (see run history for this commit). v3.27.0 switched to using foreground cascading deletion, which is causing our stacks to fail deletion on finalization. I'm pinning to v3.26.0 for now to unblock CI and releases, until we can determine a better solution going forwards.
This PR addresses 3 outstanding issues with the current broken tests:
Ordered
decorator.