The e2e-metal job relies on loadbalancer service that was set up manually on the build01 cluster.

The service and any related resources should instead be managed in https://github.com/openshift/release/tree/master/clusters/build-clusters.

Additionally, if the service can only be set up on one build cluster, then the job must be pinned to that cluster. However, DPTP has suggested that we should only pin as a last resort. See https://github.com/openshift/release/blob/941ca83ec8c6832680b73cf4f94eb1c0beaac9f1/core-services/sanitize-prow-jobs/_config.yaml#L3944-L3956.

Slack discussion: https://coreos.slack.com/archives/CHY2E1BL4/p1615579803083100