See, there's a bunch of stuff that I deploy using terraform (VPC, DNS, EKS) and a bunch of stuff that I deploy with FluxCD. But in between, there's some awkward stuff like the monitoring stack that requires cloudy things and Kube things, and they're tightly coupled.
Right now I end up going with terraform and the awful helm provider. Many of these helm charts have sub-charts, nested values etc, but thankfully the monitoring stack doesn't change that much. It's still not ideal but it works as a one-shot deployment that sets up buckets, policies, IAM roles for service accounts, more policies for the roles, and finally the helm charts with values from the outputs of those clouds resources.