Coming back to something from earlier in the thread, because it keeps being talked past:
I disagree that testing history is decisive. It tells you what a supplier did when they were being watched. Continuity of behaviour under stress — a late shipment, a failed test, a complaint — is more predictive than any run of good results.