Description
Landing-12 review M1 (codex/loop58/rv12/review.md): after SKT-0200, requests failed only by the entry’s baseline error draw keep the legacy span nesting (D85), so error traces in healthy windows show untouched services about 4x slower than success traces, and learnHealthySpans adds their hop durations to the CLIENT/SERVER healthy references, biasing faulted-row factors low by about error_rate x 3. Must be fixed before a shipped blueprint opts in.
Acceptance Criteria
- #1 Baseline-failed causal requests use the engine timeline while keeping their drawn status and error code
- #2 Healthy span references exclude any request not laid out on the engine timeline, and faulted-row factors match the engine factor within sampling noise at 1% and 5% error rates
Definition of Done
- #1 just check (fmt-check, lint, gen-check, env-check, docs-check, test, race, hygiene, ui-check, compose-check, helm-test, lab-check, signal-fidelity)
- #2 just gen (only if a blueprint field, construct/workload config struct, or a skill under plugins/synthkit/skills/ changed)
- #3 just dump — inventory diffed against signals/
Implementation Notes
loop58 closeout (2026-10-25): parked unlanded; see /Users/rob/repos/synthkit/codex/closeout-2026-10-25-loop58.md. Accepted and in landing-13 mbox (codex/loop58/archive/land13-d4a51b34..3c664fc6.mbox); landing blocked by SKT-0211 (amd64 FMA digest drift). Resume: fix SKT-0211, re-gate, land.
Owner closeout: accepted prior timing/join candidate remains in the retained archived causal stack, not in published database-only delivery. Resume from the explicit engine-only split with accepted timestamp/resend fixes, preserving existing assertions and excluding the held reference migration; final composition, generation, review and exact-tree gate remain required.
Loop60 closeout: Accepted task-separated candidate is retained in the repaired fe995479 causal stack, not published. Composed app race still exceeds the unchanged600-second budget, including after exact optimization foundation; plain/component success is not full-gate acceptance. Resume after owner-authorized evaluator resolution, current-main composition, final generation/review/full gate and actual inventory proof. No timeout/assertion/population weakening or unchanged green sampling.
Final loop closeout: retained unlanded causal candidateaf34 has native/component review and final CodeRabbit passes, but the exact integrated full gate fails app-race600.139s. No candidate source was pushed or deployed. Preserve task-specific earlier evidence/limits; resume only after the aggregate owner stop and relevant dependency/owner decisions are resolved, with no timeout/assertion/case/isolation reduction or unchanged full-gate retry.