# E111 Six-Hour Persistent Governed Mission

- Run directory: `/root/cain-e110-infra/artifacts/e111/endurance`
- Label: **SIX_HOUR_RESUME**
- State at bundle build: **DONE**
- Started (epoch): `1795128502.4647756`
- Deadline (epoch): `1795133326.4647756`
- Observed journal span: **21923.5 s** over **42,052** hash-chained tasks
- Verified solved: **36,045**; failed: **6007**;
  false accepts: **0**; verticals: **{'software': 6008, 'security': 6008, 'mathematics': 6008, 'science': 12014, 'data': 6007, 'business': 6007}**
- Journal chain verifies: **True**
- Published summary: `{"label": "SIX_HOUR_RESUME", "hours_requested": 1.34, "elapsed_s": 4824.156, "journal_span_s": 21923.475, "resumed_from_seq": 32798, "completed": true, "tasks": 9254, "duplicates": 0, "journal_seq": 42052, "journal_head": "88df720eda69f76769c52582442c3b5986bb54490f1b70bd3b6a0e22ccb3208e", "journal_ok": true, "peak_rss_mb": 127.7, "permits_used": 0, "killed": false, "status": "PASS", "stop_elapsed_s": 16815.338, "cumulative_s": 21639.494}`
- **Two segments:** the first segment was stopped on owner instruction at **4.67 h**;
  the second resumed for **1.34 h**. Cumulative active execution
  **6.01 h**. Because of the stop, the continuous six-hour criterion is **not** met: G1 is
  **PARTIAL**.
- **Separate one-hour run:** **PASS** — one-hour run: 3600.257 s, 6957 tasks, status PASS, journal_ok True.
- **Known defect:** all failed tasks are the same mission (`{'M4_science': 6007}`), caused by a
  defective E110 `M4_science` verifier that required an interventional table it never received. This is a harness
  defect, not an infrastructure failure; it is disclosed and not counted as a success.

The mission runs deterministic, verifiable integration missions across software, security, mathematics, science,
data, business and an abstention case, round-robin, checkpointing every task. Each task is appended to a
SHA-256 chain; a duplicate task id is refused, not re-executed.

## Interruption test (separate short run, same runner)

A SIGKILL was delivered after 25 tasks; the runner restarted on the same journal and resumed at sequence 26.
Events before kill: **25**; after restart: **342**;
duplicate task ids: **0**; chain: **True**;
recovery: **1.001 s**.

## Honest limit

The six-hour process was already live when the interruption test was scheduled, so the kill was performed on a
separate short run of the same code, not inside the live six-hour run. Coordinator restart, revocation survival
and the KILL fence are demonstrated in the independence artifact. If the run has not reached 21,600 s at build
time, G1 is **PARTIAL**, never PASS.

## Resource accounting

Peak RSS (sampled): **37992** KB; CPU seconds between first and last sample:
**17.23**; throughput:
**1.9181** tasks/s.
