fix(alerts): Steward's own outage no longer reads as every agent going silent (#5441)

After Steward or its database was down, every agent's age and every
series' last sample were old, so all agent-offline rules fired and all
rules went no-data at once, clearing a minute later. With per-account
alert email that would have mailed everyone a dozen notices per restart.

A periodic job now ticks a Watch (steward/core/watch.py) once it has
reached the database; the Watch says since when runs have been unbroken
(process start, or the end of a gap over 120 s). seconds_since_seen and
the no-data sweep count silence from no earlier than that, so a source
must stay quiet while Steward is watching before it counts as silent.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>

tests/core/test_alert_no_data.py

@@ -60,6 +60,19 @@ def test_silent_since(last_ago, interval, watch_ago, silent_after):
         assert got == (last or watch) + timedelta(seconds=silent_after)
 
 
+def test_steward_s_own_outage_is_not_the_series_silence():
+    """#5441: a series last sampled 18 minutes ago, while Steward has only been
+    back for 2 minutes, is not silent yet; once Steward has watched it for
+    three of its gaps with nothing arriving, it is, from then."""
+    last = NOW - timedelta(minutes=18)
+    watch = NOW - timedelta(days=1)
+    assert silent_since(last, 60, watch, NOW, awake_since=NOW - timedelta(minutes=2)) is None
+    awake = NOW - timedelta(minutes=6)
+    assert silent_since(last, 60, watch, NOW, awake_since=awake) == awake + timedelta(seconds=300)
+    # A sample since Steward came back is what counts, when there is one.
+    assert silent_since(NOW - timedelta(minutes=1), 60, watch, NOW, awake_since=awake) is None
+
+
 class _Session:
     def __init__(self):
         self.added = []

tests/core/test_watch.py

@@ -0,0 +1,32 @@
+"""Since when Steward has been watching without a break (#5441): a job's
+Watch starts at its first run, holds across runs a minute apart, and starts
+again after a gap long enough to mean Steward was not watching."""
+from datetime import datetime, timedelta, timezone
+from types import SimpleNamespace as NS
+
+from steward.core.watch import GAP_S, Watch, watch
+
+T0 = datetime(2026, 10, 9, 2, 0, tzinfo=timezone.utc)
+
+
+def test_the_first_run_starts_the_watch_and_steady_runs_keep_it():
+    w = Watch()
+    assert w.tick(T0) == T0
+    for minute in range(1, 30):
+        assert w.tick(T0 + timedelta(minutes=minute)) == T0
+
+
+def test_a_gap_longer_than_two_missed_runs_starts_it_again():
+    w = Watch()
+    w.tick(T0)
+    assert w.tick(T0 + timedelta(seconds=GAP_S)) == T0          # just within
+    back = T0 + timedelta(seconds=2 * GAP_S + 1)
+    assert w.tick(back) == back                                   # Steward was away
+    assert w.tick(back + timedelta(minutes=1)) == back
+
+
+def test_each_job_has_its_own_watch_on_its_own_app():
+    a, b = NS(), NS()
+    assert watch(a, "x") is watch(a, "x")
+    assert watch(a, "x") is not watch(a, "y")
+    assert watch(a, "x") is not watch(b, "x")