This message was deleted.

I've read the tf doc a couple times now and think i'm understanding what they're saying, but yes - it's very common to bootstrap a cluster and the resources in it inside the same stack/program

the work-around in Terraform is an interesting read - for AKS at least: https://github.com/hashicorp/terraform-provider-kubernetes/blob/main/_examples/aks/README.md - but it strikes me as quite the cludge. Surely treating Providers with lifecycle similar to resources would resolve such issues.

Pulumi's provider model is different, you can set specific providers for each individual resource, it's a much cleaner model