fix(alertmanager): keep severity=none alerts non-actionable

[allyblockcast]

Summary

Paperclip issue: https://paperclip.blockcast.net/BLO/issues/BLO-24177

• Retain severity=none delivery evidence as done and unassigned.
• Recover and refresh the same terminal row after state loss.
• Clear escalation metadata and bypass owner/route resolution for terminal severities.
• Add regression coverage for creation and state-loss recovery.

Verification

• git diff --check clean.
• Focused typecheck could not run in the fresh clone because dependencies are not installed (tsc: not found).

Acceptance criteria

• Watchdog rows remain unassigned and terminal.
• Delivery evidence remains refreshable.
• Normal severities retain existing routing behavior.

[allyblockcast]

🔗 Paperclip issue: BLO-24177

[allyblockcast]

🔗 Paperclip issue: BLO-24177

[allyblockcast]

Hey @allyblockcast[bot]! Before this PR can be reviewed, a few things need attention:

Missing or incomplete:

• Missing section: ## Thinking Path
• Missing section: ## What Changed
• Missing section: ## Risks
• Missing section: ## Model Used
• Add the dedup-search checkbox to your PR description and check it once you have searched the GitHub PR list for similar PRs. See the PR template at .github/PULL_REQUEST_TEMPLATE.md and CONTRIBUTING.md → "Before You Start: Search First".

Once updated, push a new commit and these checks will re-run automatically.

— commitperclip

[allyblockcast]

Ally — Consolidated PR Review

Lenses: pr-review-toolkit (code, tests, comments, errors, types) + gstack/review + native-codex.
Reviewed head: fd85c3e

Critical Issues (0)

Important Issues (1)

• [native-codex] packages/plugins/paperclip-plugin-alertmanager/src/webhook-handler.ts:790 — terminal re-fires return before aggregate membership and firing-event reconciliation. When a terminal alert fires, resolves, and later fires again, this branch clears resolvedAt in state and refreshes the issue, but skips upsertAggregateMember and alertmanager.alert.firing emission. The aggregate member can remain marked resolved, so a later resolve may close the shared aggregate issue despite this fingerprint firing, and downstream consumers miss the re-fire.
  • Run terminal re-fires through the same aggregate fence/member upsert and event/state update path as normal re-fires, or explicitly reset the member and emit the firing event before returning. Add a regression test covering terminal -> resolved -> terminal re-fire and asserting membership is unresolved and the firing event is emitted.

Suggestions (0)

Strengths

• The terminal severity predicate is isolated, normalized case-insensitively, and supports an explicit replacement override.
• Owner and route resolution are avoided for newly created terminal alerts, matching the non-actionable requirement.
• State-loss recovery is covered for an existing terminal issue.

Recommended Action

1. Fix the Important issue before merge.
2. Add the terminal recurrence regression coverage.