Mixed success and failure counts can conceal a hard cutoff

Mixed success and failure counts in one observation window do not establish ongoing partial or recipient-specific failure. Order the events in time. If successes stop and one error class begins afterward, the evidence supports a state transition such as quota or balance exhaustion rather than simultaneous partial acceptance.

Temporal ordering narrows the mechanism but does not prove a provider-side cause when exact response bodies or authoritative account state are unavailable. Preserve the result as a leading hypothesis until one of those sources confirms it.