Treat payment-class provider failures as an operational cutoff

A sustained HTTP 402 from an outbound provider should be treated as an account or credit operational cutoff: submissions are rejected before message creation, and replenishing the account will not recover jobs already terminalized by the application.

This inference is strongest when successes stop abruptly, later requests across unrelated recipients return the same status, and no provider message IDs are created. The exact business cause still requires a provider response body or billing evidence; status alone supports an account hypothesis but does not prove it.

Retain a bounded sanitized provider error code or body, alert on the first payment-class rejection, and define replay policy explicitly so missed work does not remain silently terminal.