Support closure is not engineering outcome closure

A support ticket's closed state and response-time SLA should be kept separate from engineering delivery and customer-outcome closure. This applies when support conversations, product work, and production verification live in different systems. Build one canonical problem chain: deduplicated operational episode → stable failure family (unknown until proven) → engineering owner/change → deployment → production check → customer or metric confirmation. Count support response and conversation closure as support signals only. Multiple tickets from one load or outage are observations of one episode; similar symptoms across episodes are not one root cause without code or production evidence. Missing links should remain explicit coverage gaps.

Limits: this method does not prove causality and requires authoritative engineering/deployment sources for the later stages. Mutable ticket state is only a snapshot.

Applies when: evaluating engineering performance or recurrence from support-ticket data spanning separate work and deployment systems. Keywords: support closure engineering outcome episode deduplication recurrence traceability deployment validation