While investigating some intermittent GCP CI errors, it was determined that the failure was due to a required GCP API service being disabled. This information was in the CCO Pod logs, but could have been bubbled up to the CredentialsRequest status (and possibly the ClusterOperator status too).

See https://docs.google.com/document/d/1iQpyO-e1Z7SgZdGKrp8PR6ITA3C7BG1KgOEgTbIH3ig/edit# for a full writeup.