fix: #412/#413/#414 runtime stability and PR governance preflight

## #413 — runtime_verify_monitor.sh pipefail fix
- find_live_pids() now captures pgrep output via local variable with || true
  so pipefail never triggers on no-match (pgrep exit 1)
- Regression test added: monitor survives MAX_LOOPS=1 without crash

## #412 — startup crash logging improvement
- Add asyncio.CancelledError catch in sync_positions_from_broker call
  so BaseException-level cancellations are logged before propagating
- Provides evidence in run log if CancelledError causes future startup aborts

## #414 — PR governance preflight
- validate_pr_body.py: add REQ-ID/TASK-ID/TEST-ID pattern checks (--no-governance flag to skip)
- docs/workflow.md: new "PR Governance Preflight (Mandatory)" section
- docs/commands.md: "PR Body Governance Preflight" section before tea pulls create
- Tests: 4 new governance traceability tests in test_validate_pr_body.py

Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>

tests/test_runtime_overnight_scripts.py

@@ -158,3 +158,46 @@ def test_run_overnight_fails_when_process_exits_before_grace_period(tmp_path: Pa
         watchdog_pid = int(watchdog_pid_file.read_text(encoding="utf-8").strip())
         with pytest.raises(ProcessLookupError):
             os.kill(watchdog_pid, 0)
+
+
+def test_runtime_verify_monitor_survives_when_no_live_pid(tmp_path: Path) -> None:
+    """Regression test for #413: monitor loop must not exit when pgrep finds no live process.
+
+    With set -euo pipefail, pgrep returning exit 1 (no match) would cause the
+    whole script to abort via the pipefail mechanism. The fix captures pgrep
+    output via a local variable with || true so pipefail is never triggered.
+
+    Verifies that the script: (1) exits 0 after completing MAX_LOOPS=1, and
+    (2) logs a HEARTBEAT entry. Whether live_pids is 'none' or not depends on
+    what processes happen to be running; either way the script must not crash.
+    """
+    log_dir = tmp_path / "overnight"
+    log_dir.mkdir(parents=True, exist_ok=True)
+
+    env = os.environ.copy()
+    env.update(
+        {
+            "ROOT_DIR": str(REPO_ROOT),
+            "LOG_DIR": str(log_dir),
+            "INTERVAL_SEC": "1",
+            "MAX_HOURS": "1",
+            "MAX_LOOPS": "1",
+            "POLICY_TZ": "UTC",
+        }
+    )
+    completed = subprocess.run(
+        ["bash", str(RUNTIME_MONITOR)],
+        cwd=REPO_ROOT,
+        env=env,
+        capture_output=True,
+        text=True,
+        check=False,
+    )
+    assert completed.returncode == 0, (
+        f"monitor exited non-zero (pipefail regression?): {completed.stderr}"
+    )
+    log_text = _latest_runtime_log(log_dir)
+    assert "[INFO] runtime verify monitor started" in log_text
+    assert "[HEARTBEAT]" in log_text, "monitor did not complete a heartbeat cycle"
+    # live_pids may be 'none' (no match) or a pid (process found) — either is valid.
+    # The critical invariant is that the script survived the loop without pipefail abort.

tests/test_validate_pr_body.py

@@ -37,6 +37,7 @@ def test_validate_pr_body_text_allows_escaped_newline_in_code_blocks() -> None:
         [
             "## Summary",
             "- example uses `\\n` for explanation",
+            "- REQ-OPS-001 / TASK-OPS-001 / TEST-OPS-001",
             "```bash",
             "printf 'line1\\nline2\\n'",
             "```",
@@ -63,7 +64,7 @@ def test_validate_pr_body_text_passes_with_valid_markdown() -> None:
     text = "\n".join(
         [
             "## Summary",
-            "- item",
+            "- REQ-OPS-001 / TASK-OPS-001 / TEST-OPS-001",
             "",
             "## Validation",
             "```bash",
@@ -74,6 +75,34 @@ def test_validate_pr_body_text_passes_with_valid_markdown() -> None:
     assert module.validate_pr_body_text(text) == []
 
 
+def test_validate_pr_body_text_detects_missing_req_id() -> None:
+    module = _load_module()
+    text = "## Summary\n- TASK-OPS-001 / TEST-OPS-001 item\n"
+    errors = module.validate_pr_body_text(text)
+    assert any("REQ-ID" in err for err in errors)
+
+
+def test_validate_pr_body_text_detects_missing_task_id() -> None:
+    module = _load_module()
+    text = "## Summary\n- REQ-OPS-001 / TEST-OPS-001 item\n"
+    errors = module.validate_pr_body_text(text)
+    assert any("TASK-ID" in err for err in errors)
+
+
+def test_validate_pr_body_text_detects_missing_test_id() -> None:
+    module = _load_module()
+    text = "## Summary\n- REQ-OPS-001 / TASK-OPS-001 item\n"
+    errors = module.validate_pr_body_text(text)
+    assert any("TEST-ID" in err for err in errors)
+
+
+def test_validate_pr_body_text_skips_governance_when_disabled() -> None:
+    module = _load_module()
+    text = "## Summary\n- item without any IDs\n"
+    errors = module.validate_pr_body_text(text, check_governance=False)
+    assert not any("REQ-ID" in err or "TASK-ID" in err or "TEST-ID" in err for err in errors)
+
+
 def test_fetch_pr_body_reads_body_from_tea_api(monkeypatch) -> None:
     module = _load_module()