feat: a tick keeps looking back 30 days, so an EDITED post is reached (4386)

Operator, 2026-09-23, on a Floppystack post: "this post has been updated as
he implements hot fixes — any chance we have a way to scan for or see updated
posts so we can update ours to match and pull the new attachments and
pictures etc."

The download half already worked: extract_media reads the media list off the
LIVE feed response every walk, so a newly attached hotfix build is a ledger
key we have never seen. Only REACHING the post was missing — a tick stopped
after 20 contiguous already-have-it items, and a post edited three days after
publication sits well below twenty. Not a bug in the early-out; a count
cannot express "recent".

The early-out now needs BOTH conditions: the run of seen items AND a post
published before the horizon. Strictly a widening — window 0 is exactly the
old behaviour, and no window can make a tick stop EARLIER than it used to, so
a source paused for months still walks its whole unseen backlog. The horizon
is a floor on how far to look, never a ceiling.

Inside the window the post-record gate is bypassed too (write_post_record
revisit=True): the body is re-read from the feed response already in hand, so
a revisit costs zero requests, and a body that comes back empty writes
NOTHING rather than blanking one a detail-fetch had filled. Revisits are kept
out of the #862 body-drift canary's sample for the same reason — an empty
revisit is healthy, and counting it would walk the alarm toward firing on
good ticks.

The run summary names what changed ("3 post(s) updated (5 new file(s))") with
a line per post; the ask was to SEE updated posts, not only to end up with
their bytes.

download_revisit_days is a settings row, not a constant (rule 25) — how long
a creator keeps editing is a property of the creator. Default 30, 0 turns it
off. Migration 0108.

Also corrects two stale docstrings: both clients described post_meta as
feeding an Ingester.preview that no longer calls it. It had no consumer at
all until this change.

Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01LVjrnpQjRgHdvq95rASoiR

tests/test_patreon_ingester.py

@@ -7,6 +7,8 @@ real CDN. The ledger is real (a sync sessionmaker bound to the test engine), so
 the tier-1 skip and the idempotent mark-seen run against actual rows.
 """
 
+from datetime import UTC, datetime, timedelta
+
 import pytest
 from sqlalchemy import func, select
 from sqlalchemy.orm import sessionmaker
@@ -50,9 +52,15 @@ class _FakeClient:
     """Stub PatreonClient. `pages` is a list of (page_cursor, [posts]); each post
     is (post_id, [MediaItem]). `raise_on_first` lets a test trip drift."""
 
-    def __init__(self, pages, raise_exc=None, empty_body=False, gated=None):
+    def __init__(self, pages, raise_exc=None, empty_body=False, gated=None,
+                 published=None):
         self._pages = pages
         self._raise_exc = raise_exc
+        # {post_id: ISO-8601 published_at} for the revisit window. Absent → the
+        # date reads as None, which is how EVERY test written before the window
+        # existed keeps its old behaviour: an unreadable date is never inside
+        # the horizon, so the count early-out stands alone.
+        self._published = dict(published or {})
         # empty_body simulates a body-field schema break: every post comes back
         # with no content (the #862 canary's trip condition).
         self._empty_body = empty_body
@@ -89,7 +97,10 @@ class _FakeClient:
         return post["_media"]
 
     def post_meta(self, post):
-        return {"title": post.get("id"), "date": None}
+        return {
+            "title": post.get("id"),
+            "date": self._published.get(str(post.get("id") or "")),
+        }
 
     @staticmethod
     def post_is_gated(post):
@@ -113,6 +124,7 @@ class _FakeDownloader:
         self.error = set(error or ())
         self.download_calls = 0
         self.post_records = 0
+        self.post_revisits = 0
 
     def download_post(self, post, media_items, artist_slug, *, is_seen,
                       should_stop=lambda: False, recapture=False):
@@ -146,8 +158,19 @@ class _FakeDownloader:
                 outcomes.append(MediaOutcome(media=m, status="downloaded", path=p, error=None))
         return outcomes
 
-    def write_post_record(self, post, artist_slug):
+    def write_post_record(self, post, artist_slug, *, revisit=False):
         self.post_records += 1
+        if revisit:
+            self.post_revisits += 1
+        attrs_ = post.get("attributes") or {}
+        body_ = attrs_.get("content")
+        # Mirrors the real downloaders' revisit contract: a re-read whose body
+        # came back empty writes NOTHING rather than blanking a stored one.
+        if revisit and not (isinstance(body_, str) and body_.strip()):
+            return PostRecordOutcome(
+                path=None, post_type=attrs_.get("post_type"),
+                title=attrs_.get("title"), body_chars=0,
+            )
         p = self.tmp_path / f"{post.get('id')}__post.json"
         p.write_text("{}")
         attrs = post.get("attributes") or {}
@@ -1114,3 +1137,289 @@ async def test_body_canary_silent_below_min_sample(source_id, sync_engine, tmp_p
     )
     assert result.success is True
     assert result.error_type is None
+
+
+# --- the revisit window: reaching posts that were EDITED after capture -------
+#
+# Operator, 2026-09-23, on a Floppystack post: *"this post has been updated as
+# he implements hot fixes — any chance we have a way to scan for or see updated
+# posts so we can update ours to match and pull the new attachments and
+# pictures etc."*
+#
+# A tick stopped after N contiguous already-have-it items, so an edit to a
+# three-day-old post was structurally unreachable: it sits well below twenty
+# seen items. The early-out now needs BOTH that run AND a post published before
+# the horizon.
+
+
+def _iso(days_ago):
+    return (datetime.now(UTC) - timedelta(days=days_ago)).isoformat()
+
+
+def _seed_seen(sync_engine, source_id, media):
+    factory = sessionmaker(sync_engine, expire_on_commit=False)
+    with factory() as s:
+        for m in media:
+            s.add(PatreonSeenMedia(
+                source_id=source_id, filehash=_ledger_key(m), post_id=m.post_id,
+            ))
+        s.commit()
+
+
+@pytest.mark.asyncio
+async def test_a_run_of_seen_items_does_not_stop_a_tick_inside_the_window(
+    source_id, sync_engine, tmp_path,
+):
+    """The bug itself. Three all-seen posts, threshold 2 — the old walk turned
+    around at the second and never looked at the third, which is exactly where
+    an edited post lives."""
+    seen = [_media(f"p{i}", 1) for i in range(1, 4)]
+    _seed_seen(sync_engine, source_id, seen)
+
+    client = _FakeClient(
+        [(None, [(m.post_id, [m]) for m in seen])],
+        published={m.post_id: _iso(3) for m in seen},
+    )
+    ing = _ingester(sync_engine, tmp_path, client, _FakeDownloader(tmp_path))
+
+    result = ing.run(
+        source_id=source_id, campaign_id="c1", artist_slug="ingest",
+        url="https://patreon.com/ingest", mode="tick", seen_threshold=2,
+        revisit_days=30,
+    )
+
+    assert result.success is True
+    assert client.consumed_posts == 3
+
+
+@pytest.mark.asyncio
+async def test_the_early_out_still_fires_once_the_walk_is_below_the_horizon(
+    source_id, sync_engine, tmp_path,
+):
+    """The other half, and the one that keeps a tick cheap. Same three posts,
+    published outside the window → the count early-out stands exactly as it
+    did. Asserted beside the test above because the window is only correct if
+    BOTH conditions are required; either one alone is a different feature."""
+    seen = [_media(f"p{i}", 1) for i in range(1, 4)]
+    _seed_seen(sync_engine, source_id, seen)
+
+    client = _FakeClient(
+        [(None, [(m.post_id, [m]) for m in seen])],
+        published={m.post_id: _iso(90) for m in seen},
+    )
+    ing = _ingester(sync_engine, tmp_path, client, _FakeDownloader(tmp_path))
+
+    result = ing.run(
+        source_id=source_id, campaign_id="c1", artist_slug="ingest",
+        url="https://patreon.com/ingest", mode="tick", seen_threshold=2,
+        revisit_days=30,
+    )
+
+    assert result.success is True
+    assert client.consumed_posts == 2
+
+
+@pytest.mark.asyncio
+async def test_a_window_of_zero_is_the_behaviour_that_shipped_before_it(
+    source_id, sync_engine, tmp_path,
+):
+    """The off switch. Recent posts, a window of 0 → the walk stops on the
+    count alone, so an operator who wants the old cheap tick has one."""
+    seen = [_media(f"p{i}", 1) for i in range(1, 4)]
+    _seed_seen(sync_engine, source_id, seen)
+
+    client = _FakeClient(
+        [(None, [(m.post_id, [m]) for m in seen])],
+        published={m.post_id: _iso(1) for m in seen},
+    )
+    ing = _ingester(sync_engine, tmp_path, client, _FakeDownloader(tmp_path))
+
+    ing.run(
+        source_id=source_id, campaign_id="c1", artist_slug="ingest",
+        url="https://patreon.com/ingest", mode="tick", seen_threshold=2,
+        revisit_days=0,
+    )
+
+    assert client.consumed_posts == 2
+
+
+@pytest.mark.asyncio
+async def test_a_backfill_ignores_the_window_because_it_never_early_outs(
+    source_id, sync_engine, tmp_path,
+):
+    """Stated so the window cannot quietly acquire a second job. A backfill
+    walks to the bottom regardless, and `recapture` is the mode that re-reads
+    every body — a horizon there would be a third answer to a two-answer
+    question."""
+    seen = [_media(f"p{i}", 1) for i in range(1, 4)]
+    _seed_seen(sync_engine, source_id, seen)
+
+    client = _FakeClient(
+        [(None, [(m.post_id, [m]) for m in seen])],
+        published={m.post_id: _iso(90) for m in seen},
+    )
+    downloader = _FakeDownloader(tmp_path)
+    ing = _ingester(sync_engine, tmp_path, client, downloader)
+
+    ing.run(
+        source_id=source_id, campaign_id="c1", artist_slug="ingest",
+        url="https://patreon.com/ingest", mode="backfill", seen_threshold=2,
+        revisit_days=30,
+    )
+
+    assert client.consumed_posts == 3
+    assert downloader.post_revisits == 0
+
+
+@pytest.mark.asyncio
+async def test_an_edited_post_inside_the_window_downloads_its_new_attachment(
+    source_id, sync_engine, tmp_path,
+):
+    """The operator's case, end to end.
+
+    Walk one captures a post with one file. The creator then edits it to attach
+    a hotfix build. Walk two must download that file AND name the post as
+    updated — the ask was to SEE which posts changed, not only to end up with
+    their bytes.
+
+    The detection half needed nothing new: `extract_media` reads the media list
+    off the LIVE feed response every walk, so a newly attached file is simply a
+    ledger key we have never seen. Only reaching the post was missing.
+    """
+    original = _media("p1", 1)
+    client1 = _FakeClient(
+        [(None, [("p1", [original])])], published={"p1": _iso(3)},
+    )
+    ing1 = _ingester(sync_engine, tmp_path, client1, _FakeDownloader(tmp_path))
+    ing1.run(
+        source_id=source_id, campaign_id="c1", artist_slug="ingest",
+        url="https://patreon.com/ingest", mode="tick", revisit_days=30,
+    )
+
+    hotfix = _media("p1", 2)
+    client2 = _FakeClient(
+        [(None, [("p1", [original, hotfix])])], published={"p1": _iso(3)},
+    )
+    downloader2 = _FakeDownloader(tmp_path)
+    ing2 = _ingester(sync_engine, tmp_path, client2, downloader2)
+    result = ing2.run(
+        source_id=source_id, campaign_id="c1", artist_slug="ingest",
+        url="https://patreon.com/ingest", mode="tick", revisit_days=30,
+    )
+
+    assert result.files_downloaded == 1
+    assert "1 post(s) updated (1 new file(s))" in result.stdout
+    assert "post p1 — updated: 1 new file(s)" in result.stdout
+    # The post record was re-read, not skipped by the capture gate.
+    assert downloader2.post_revisits == 1
+
+
+@pytest.mark.asyncio
+async def test_a_post_below_the_horizon_keeps_its_capture_gate(
+    source_id, sync_engine, tmp_path,
+):
+    """The gate still does its job everywhere the window does not reach — else
+    the window would have quietly become "re-read every post forever"."""
+    old = _media("pold", 1)
+    client1 = _FakeClient([(None, [("pold", [old])])], published={"pold": _iso(90)})
+    ing1 = _ingester(sync_engine, tmp_path, client1, _FakeDownloader(tmp_path))
+    ing1.run(
+        source_id=source_id, campaign_id="c1", artist_slug="ingest",
+        url="https://patreon.com/ingest", mode="tick", revisit_days=30,
+    )
+
+    client2 = _FakeClient([(None, [("pold", [old])])], published={"pold": _iso(90)})
+    downloader2 = _FakeDownloader(tmp_path)
+    ing2 = _ingester(sync_engine, tmp_path, client2, downloader2)
+    ing2.run(
+        source_id=source_id, campaign_id="c1", artist_slug="ingest",
+        url="https://patreon.com/ingest", mode="tick", revisit_days=30,
+    )
+
+    assert downloader2.post_records == 0
+
+
+@pytest.mark.asyncio
+async def test_a_revisit_that_finds_nothing_new_is_not_reported_as_an_update(
+    source_id, sync_engine, tmp_path,
+):
+    """Most revisits find nothing — that is the normal case, and a run summary
+    claiming "1 post(s) updated" on every tick would train the operator to stop
+    reading it (lesson: a signal that is always on is not a signal)."""
+    m = _media("p1", 1)
+    client1 = _FakeClient([(None, [("p1", [m])])], published={"p1": _iso(3)})
+    ing1 = _ingester(sync_engine, tmp_path, client1, _FakeDownloader(tmp_path))
+    ing1.run(
+        source_id=source_id, campaign_id="c1", artist_slug="ingest",
+        url="https://patreon.com/ingest", mode="tick", revisit_days=30,
+    )
+
+    client2 = _FakeClient([(None, [("p1", [m])])], published={"p1": _iso(3)})
+    downloader2 = _FakeDownloader(tmp_path)
+    ing2 = _ingester(sync_engine, tmp_path, client2, downloader2)
+    result = ing2.run(
+        source_id=source_id, campaign_id="c1", artist_slug="ingest",
+        url="https://patreon.com/ingest", mode="tick", revisit_days=30,
+    )
+
+    assert result.files_downloaded == 0
+    assert "updated" not in result.stdout
+    assert downloader2.post_revisits == 1
+
+
+@pytest.mark.asyncio
+async def test_revisits_do_not_feed_the_body_drift_canary(
+    source_id, sync_engine, tmp_path,
+):
+    """#862's canary fails a run that recorded a meaningful sample of posts and
+    got a body from NONE of them. A revisit legitimately comes back empty — the
+    post's body only ever arrived from the detail endpoint, which a revisit
+    declines to call — so counting revisits into that sample would walk the
+    alarm toward firing on healthy ticks. First captures only."""
+    posts = [(f"c{i}", []) for i in range(_CANARY_MIN_SAMPLE)]
+    published = {pid: _iso(3) for pid, _ in posts}
+
+    client1 = _FakeClient([(None, posts)], published=published)
+    ing1 = _ingester(sync_engine, tmp_path, client1, _FakeDownloader(tmp_path))
+    first = ing1.run(
+        source_id=source_id, campaign_id="c1", artist_slug="ingest",
+        url="https://patreon.com/ingest", mode="tick", revisit_days=30,
+    )
+    assert first.success is True
+
+    # Second walk: every post is a revisit, and every body comes back empty.
+    client2 = _FakeClient([(None, posts)], published=published, empty_body=True)
+    ing2 = _ingester(sync_engine, tmp_path, client2, _FakeDownloader(tmp_path))
+    second = ing2.run(
+        source_id=source_id, campaign_id="c1", artist_slug="ingest",
+        url="https://patreon.com/ingest", mode="tick", revisit_days=30,
+    )
+
+    assert second.success is True
+    assert second.error_type is not ErrorType.API_DRIFT
+
+
+@pytest.mark.asyncio
+async def test_a_post_whose_date_will_not_parse_falls_back_to_the_count(
+    source_id, sync_engine, tmp_path,
+):
+    """A date we cannot read must not fail the walk, and must not be guessed
+    into the window. It reads as "not provably recent", which leaves that post
+    on the behaviour it had before the window existed."""
+    seen = [_media(f"p{i}", 1) for i in range(1, 4)]
+    _seed_seen(sync_engine, source_id, seen)
+
+    client = _FakeClient(
+        [(None, [(m.post_id, [m]) for m in seen])],
+        published={m.post_id: "last Tuesday" for m in seen},
+    )
+    ing = _ingester(sync_engine, tmp_path, client, _FakeDownloader(tmp_path))
+
+    result = ing.run(
+        source_id=source_id, campaign_id="c1", artist_slug="ingest",
+        url="https://patreon.com/ingest", mode="tick", seen_threshold=2,
+        revisit_days=30,
+    )
+
+    assert result.success is True
+    assert client.consumed_posts == 2