For our E2E, we will need an ACM hub, at least one managed cluster, a KCP instance, and a cluster on which we will install the compute controller.

We will likely want to stand up our own instance of KCP rather than using kcp-stable or kcp-unstable so we can manage the version.

The compute controller can likely run on the ACM hub to reduce the cost and complexity of the test setup. 

We can look at using the prow setup in idp that would check out a cluster claim and install ACM as a starting point. Other options to consider are leveraging Zack's pipeline to stand up our stack, doing something like the singapore demo which uses kind clusters and a local KCP, and looking at what KCP is doing for e2e tests.

Dom also mentioned VCluster as an option to consider for creating the managed clusters. 

I believe we can be flexible on where we run these tests (Prow, GitHub actions, OpenShift pipelines, etc) as long as we can trigger on PRs and merges to main and get results back into GitHub.