# Runtime Health Remediation Traceability Matrix

| Failure ID | Observed condition | Root code surface | Repair order | Required proof | Current status | Evidence |
|---|---|---|---|---|---|---|
| RHR-F01 | Unit test contacted live broker | emotion/backend construction and test bootstrap | WO-00 | zero-socket focused suites | deterministic accepted | [network inventory and hermetic gate](./WO-00-INFERENCE-NETWORK-INVENTORY.md) |
| RHR-F02 | Background inference competed with foreground task | Telegram sweep, emotion admission, `activeTask` isolation | WO-01 | zero background calls under foreground lease | deterministic accepted | [production queue/preemption receipt](./WO-01-TELEGRAM-COGNITION-INTEGRATION-EVIDENCE.md); [100-offer pressure gate](./WO-07-CONTROLLED-LOAD-EVIDENCE.md) |
| RHR-F03 | Broker queue deadline caused repeated full requests | universal admission header and retry loop | WO-02 | one logical admission identity | deterministic accepted; protocol activated | [body-free resume and activation receipt](./WO-02-BROKER-PROTOCOL-EVIDENCE.md) |
| RHR-F04 | Permanent placement and retryable wait shared a policy | broker reason mapping and client retry classification | WO-02 | no-loop permanent refusal and nonfatal retryable wait | deterministic accepted; protocol activated | [typed placement and retry evidence](./WO-02-BROKER-PROTOCOL-EVIDENCE.md) |
| RHR-F05 | Workboard rejected present evidence | prose substring matcher | WO-03 | typed predicate acceptance tests | deterministic accepted | [typed workboard evidence](./WO-03-WORKBOARD-FAILURE-LIFECYCLE-EVIDENCE.md) |
| RHR-F06 | Rejected completion activated later phases | runner phase advancement | WO-03 | dependency-safe sequence replay | deterministic accepted | [frontier and lifecycle replay](./WO-03-WORKBOARD-FAILURE-LIFECYCLE-EVIDENCE.md) |
| RHR-F07 | Repeated diagnostics grew without bound | workboard diagnostic event reducer | WO-03 | 10,000-repeat cardinality test | deterministic accepted | [10,000-event bound](./WO-03-WORKBOARD-FAILURE-LIFECYCLE-EVIDENCE.md) |
| RHR-F08 | Generic shell activity counted as verification | runner verification shortcut | WO-03, WO-05 | false-positive verifier regression | deterministic accepted | [typed verifier evidence](./WO-03-WORKBOARD-FAILURE-LIFECYCLE-EVIDENCE.md); [completion authority](./H6-MAIN-TUI-WO-05-CONFIRMATION-EVIDENCE.md) |
| RHR-F09 | Failure cards survived later success | failure identity/store resolution split | WO-03 | atomic resolution receipt test | deterministic accepted | [recoverable cross-store receipt](./WO-03-WORKBOARD-FAILURE-LIFECYCLE-EVIDENCE.md) |
| RHR-F10 | Todo mirror retained stale/reopened cards | partial todo-to-workboard reconciliation | WO-03 | removal and no-reopen tests | deterministic accepted | [lifecycle replay](./WO-03-WORKBOARD-FAILURE-LIFECYCLE-EVIDENCE.md) |
| RHR-F11 | Mutable task state appeared more than once | startup state and turn-frame overlap | WO-04 | duplicate-free context replay | deterministic accepted | [canonical projection replay](./WO-04-CANONICAL-PROJECTION-EVIDENCE.md) |
| RHR-F12 | Raw tool and run-evidence bodies duplicated | context engine plus raw transcript | WO-04 | one representation per transaction | deterministic accepted | [receipt/artifact replacement evidence](./WO-04-CANONICAL-PROJECTION-EVIDENCE.md) |
| RHR-F13 | Selector and transport used different caps | frame selection then system-message truncation | WO-04 | final-byte trace parity | deterministic accepted | [selection/transport parity](./WO-04-CANONICAL-PROJECTION-EVIDENCE.md) |
| RHR-F14 | Tool schemas consumed about 50 KB | broad core set and sticky promotion | WO-04 | schema budget and TTL tests | deterministic accepted | [schema budget and promotion TTL](./WO-04-CANONICAL-PROJECTION-EVIDENCE.md) |
| RHR-F15 | Completion could finalize with open work | advisory closeout and weak receipt | WO-05 | host readiness rejection test | deterministic accepted | [completion-readiness confirmation](./H6-MAIN-TUI-WO-05-CONFIRMATION-EVIDENCE.md) |
| RHR-F16 | Telegram active state could hide poll failure | missing poll health telemetry | WO-06 | empty-success versus error status test | deterministic accepted | [poll-health continuity receipt](./H1-H2-H7-TELEGRAM-CONTINUITY-EVIDENCE.md) |
| RHR-F17 | Telegram offset advanced before durable intake | in-memory offset and append-only log | WO-06 | crash-before-dispatch replay test | deterministic accepted | [spool/checkpoint replay receipt](./H1-H2-H7-TELEGRAM-CONTINUITY-EVIDENCE.md) |
| RHR-F18 | SQLite and handler errors could be hidden | no-op persistence/catch-and-continue paths | WO-06 | surfaced storage/handler failure tests | deterministic accepted | [failure and shutdown receipt](./H1-H2-H7-TELEGRAM-CONTINUITY-EVIDENCE.md) |
| RHR-F19 | Exit/resume could strand bridge lifecycle | non-awaited exits and resume skip | WO-06 | shutdown and resume lifecycle tests | deterministic accepted | [poll replacement and awaited-close receipt](./H1-H2-H7-TELEGRAM-CONTINUITY-EVIDENCE.md) |
| RHR-F20 | Cross-system regressions escaped focused tests | missing integrated replay/load/canary gate | WO-07 | deterministic aggregate and load thresholds, then rollback, device, operator, and live canaries | deterministic aggregate/load/shadow/rollback accepted; physical-device/operator/live pending | [controlled-load receipt](./WO-07-CONTROLLED-LOAD-EVIDENCE.md); [shadow/rollback receipt](./WO-07-SHADOW-ROLLBACK-EVIDENCE.md); [CLI aggregate](./CLI-CURRENT-MAIN-SUITE-EVIDENCE.md); [clean build](./WO-19-CLEAN-BUILD-EVIDENCE.md); [aggregate audit](./AUDIT.md) |

## Requirement-to-artifact policy

Every completed row links to a deterministic test or operational receipt.
A commit hash alone is not proof. A model statement is not proof. Live logs may
support a canary only after the hermetic and deterministic gates pass. A status
that names pending rollout or live work is intentionally not a completion claim.
