Task · CFO-0079.03

Epoch-fenced per-collector attribution observation helper

Description

Loop21 existing first-poll helper lacks request and limiter-wait attribution. A private bounded read-only helper is prerequisite to the authorised measure-first live table; no upstream Cloudflare calls, deploy or throughput change.

Acceptance Criteria

Definition of Done

Implementation Plan

Reuse inspected first-poll matrix judging and credential convention. Add private bounded helper with explicit metric inputs for M1 identifiers; authentic offline malformed/missing/reset/gap fixtures, no live calls from lane. Root drives observation only after deployment.

Implementation Notes

Loop21 helper attempt1/reviewround1: offline gate2tests green but reviewer adverse CLI proves false epoch fence and demand normalization: API samples omitted from source boundary. Candidate rejected; repair adds five API instruments to boundary/preexisting checks and two authentic regressions.

Loop21: helper attempts2,reviewrounds2; private offline CLI4tests pass, independent delta review closes epoch/API boundary false-pass across all5instruments. CodeRabbit skipped ignored helper, no coverage claimed. No committed helper SHA: source hashba4bc03f6a5763a2eb7afa84c8ff7a2caafb22c165d99cebe64b71273aee7be6; observe839781747a57029435e5b2fa3612213757d19a26a8e8f0d6f5eaf13baef1d7f6. Root-only live measurement uses confirmed metric inputs; ingestion/overlap/restart limitations remain explicit.

Final Summary

Private helper prerequisite complete by authentic offline CLI regression and independent review. No runtime code or live health acceptance claimed; no packaging or signal declarations added.

View the source file on GitHub