[mashkovd]

Refs #1280, mctlhq/mctl-agents#195. Owner gate. This changes a declared measurement constraint, recorded before any producer emits, as the ADR's soak declaration requires.

Why

The soak declared "real producer traffic: none", and any producer must be recorded as an amendment before it lands. mctl-agents#195 closes only on a real DevLoop trace, and #1280's own preconditions say final scoring must use correctly exported live telemetry.

What changes

A new ADR section, "Producer amendment: DevLoop traces", plus two pointers in the runbook. Window, synthetic volume, storage, retention and teardown date are unchanged.

• Producer: the mctl-agents DevLoop only.
  • The workers carry MCTL_TRACE_WORKFLOW_TYPES=DevLoopWorkflow, so reconcile, incidents, sweep, poller and schedules record nothing.
  • The investigate and implement pods carry MCTL_TRACE_REQUIRE_PARENT=true, so manual or sweep submits stay inert.
  • mctl-agent, mctl-api and every other base-service release stay off.
• Volume: 101 investigations in the 14 days to 2026-10-10, about 7 a day.
  • Upper estimate per run: 1k spans for an investigation, 2k for an implementation.
  • That gives about 21k spans a day, or about 300k for the rest of the window.
  • Declared cap: 500,000 producer spans.
• Stop rule: any one of these turns the producer off with a one-line revert of the workers' otel.enabled:
  • the cap is reached;
  • tempo_discarded_spans_total increases;
  • either Tempo exporter alert fires;
  • Tempo's working set passes 80% of its 1Gi limit;
  • the leak check on the first trace finds a single hit.
• Not an adoption of Tempo, not a wider tracing rollout, not a window extension. The teardown commit also removes the producer env.

tests/test_adr_rubric_frozen.py passes.

🤖 Generated with Claude Code

https://claude.ai/code/session_013zqBqP9sSbVTd1kXRLA6AU

[Copilot]

Copilot was unable to review this pull request because the user who requested the review has reached their quota limit.

[github-actions]

Antigravity (agy) review — informational

Model: gemini-3.8-flash-medium · commit b6fd136.

No significant issues found.

[claude]

Claude finished @mashkovd's task in 1m 30s —— View job

---

Review complete

• Gather context (diff, CLAUDE.md, ADR spike(observability): run the live tracing backend bake-off and decide #1280)
• Review ADR amendment content
• Review runbook pointer changes
• Post findings / verdict

Verdict: APPROVE (no P1/P2 findings)

This is a docs-only change: a new "Producer amendment: DevLoop traces" section in docs/adr/0001-agent-execution-trace-backend.md plus two pointers in docs/runbooks/tracing-bake-off.md. No Helm values, ArgoCD specs, RBAC, or workflow templates touched.

Checked for correctness against actual repo config:

• MCTL_TRACE_WORKFLOW_TYPES and MCTL_TRACE_REQUIRE_PARENT are real env vars wired in platform-gitops/helm-charts/base-service/templates/deployment.yaml / rollout.yaml / _helpers.tpl, consistent with the amendment's description.
• tempo_discarded_spans_total / Tempo exporter alerts referenced in the stop rule match platform-gitops/infra-components/observability/vm-rules/tempo-eval-alerts.yaml.
• The amendment is additive (inserted before the ## Decision section) and doesn't touch the frozen rubric table or `` marker that tests/test_adr_rubric_frozen.py checks — confirmed by inspection that the new content contains no `| ... | ... |` table rows that could collide with the rubric-weights regex.
• Window, synthetic volume, storage, retention, and teardown date are unchanged as claimed; the new cap (500,000 producer spans) and stop rule are clearly declared before any producer traffic, matching the owner-gate requirement in the PR description.

Submitted: gh api .../pulls/1788/reviews with event=APPROVE, bound to commit b6fd136e.

[claude]

Docs-only ADR amendment and runbook pointers; env vars (MCTL_TRACE_WORKFLOW_TYPES, MCTL_TRACE_REQUIRE_PARENT) and metrics referenced match existing chart/alert config, no manifest or logic changes.