feat(download): smarter backfill — time-boxed chunks, run-until-done (backend)

Plan #693. Large-catalog backfill (Anduo) no longer sprints to the timeout
wall and dies as an error each run. Builds on the cursor checkpoint (#689).

- Time-boxed chunks: BACKFILL_TIMEOUT_SECONDS(1170)→BACKFILL_CHUNK_SECONDS(600),
  far under the 1350 soft limit. Hitting it = normal chunk boundary (the
  TimeoutExpired path already captures partial output + the cursor), not a
  near-wall death.
- Run-until-done state machine driven by config_overrides[_backfill_state]
  (running/complete/stalled). A running backfill auto-continues in chunks
  across ticks until gallery-dl exits cleanly (rc=0 = reached the bottom →
  'complete'); a safety-cap (BACKFILL_MAX_CHUNKS=200) + the #689 stall-guard
  pause a pathological walk as 'stalled'. Replaces the N-runs counter
  (backfill_runs_remaining repurposed as the cap countdown).
- Progress, not error: a chunk that timed out but advanced (cursor moved
  and/or files written) is reclassified TIMEOUT→PARTIAL (status 'ok').
- Retry storm tamed: gallery-dl retries 3→2, downloader timeout 120→60s, so
  one stuck CDN file fails in ~1-2 min not ~10 (Anduo #40838).
- API: POST /sources/{id}/backfill now takes {action: start|stop}; service
  start_backfill/stop_backfill; new enabled sources auto-arm run-until-done;
  source dict exposes backfill_state + backfill_chunks.

Frontend (Start/Stop control + state badge) lands in the next push.

Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>

backend/app/api/sources.py

@@ -122,26 +122,25 @@ async def delete_source(source_id: int):
 
 @sources_bp.route("/<int:source_id>/backfill", methods=["POST"])
 async def set_backfill(source_id: int):
-    """Plan #544: arm a source for backfill mode for the next N download
-    runs. Body: `{"runs": int}` (1..10, default 3). Returns the updated
-    source dict. While backfill_runs_remaining > 0, downloads use
-    gallery-dl's full-walk config (skip: True + 30-min timeout) instead
-    of the catch-up default (skip: "exit:20" + 14.5-min timeout)."""
+    """Plan #693: start or stop a run-until-done backfill. Body:
+    `{"action": "start" | "stop"}` (default "start"). 'start' walks the full
+    post history in time-boxed chunks until it reaches the bottom (then the
+    source shows 'complete'); 'stop' cancels back to tick mode. Returns the
+    updated source dict (incl. backfill_state / backfill_chunks)."""
     payload = await request.get_json(silent=True) or {}
-    runs = payload.get("runs", 3)
-    try:
-        runs = int(runs)
-    except (TypeError, ValueError):
-        return _bad("invalid_runs", detail="runs must be an integer")
+    action = payload.get("action", "start")
+    if action not in ("start", "stop"):
+        return _bad("invalid_action", detail="action must be 'start' or 'stop'")
     async with get_session() as session:
         try:
-            record = await SourceService(session).set_backfill_runs(
-                source_id, runs,
+            svc = SourceService(session)
+            record = (
+                await svc.start_backfill(source_id)
+                if action == "start"
+                else await svc.stop_backfill(source_id)
             )
         except LookupError:
             return _bad("not_found", status=404)
-        except ValueError as exc:
-            return _bad("invalid_runs", detail=str(exc))
     return jsonify(record.to_dict())
 
 

backend/app/services/download_service.py

@@ -26,8 +26,8 @@ from sqlalchemy.orm import joinedload
 from ..models import Artist, DownloadEvent, Source
 from .credential_service import CredentialService
 from .gallery_dl import (
+    BACKFILL_CHUNK_SECONDS,
     BACKFILL_SKIP_VALUE,
-    BACKFILL_TIMEOUT_SECONDS,
     TICK_SKIP_VALUE,
     ErrorType,
     GalleryDLService,
@@ -108,23 +108,20 @@ class DownloadService:
         self.sync_session.close()
 
         source_config = SourceConfig.from_dict(ctx["config_overrides"] or {})
-        # alembic 0031 / plan #544: derive skip_value + timeout from the
-        # source's backfill_runs_remaining counter. When > 0, walk the full
-        # post history (skip: True + 1170s); when 0, exit gallery-dl after
-        # 20 contiguous archived items (skip: "exit:20" + the default
-        # 870s). Operator sets backfill via POST /api/sources/{id}/backfill.
+        # Backfill mode (plan #693): the source's `_backfill_state == "running"`
+        # selects a time-boxed deep-walk chunk — skip: True (walk full history)
+        # + the BACKFILL_CHUNK_SECONDS budget, resuming from the cursor
+        # checkpoint (plan #689). State is the single source of truth; it stays
+        # "running" across ticks until the walk reaches the bottom (phase 3
+        # flips it to "complete"). Otherwise tick mode: exit gallery-dl after
+        # 20 contiguous archived items (skip: "exit:20" + the default 870s).
+        # Operator drives this via POST /api/sources/{id}/backfill {action}.
         overrides = ctx["config_overrides"] or {}
-        backfill_remaining = ctx.get("backfill_runs_remaining", 0) or 0
-        # Cursor-paged backfill (plan #689): a pending checkpoint keeps the
-        # deep-walk in backfill mode across ticks until it reaches the
-        # bottom, even after the operator's run budget hits 0 — so one large
-        # creator (Anduo) finishes without re-triggering. The phase-3
-        # stuck-guard bounds it. patreon-only: it's the sole platform with a
-        # resumable gallery-dl cursor.
-        pending_cursor = overrides.get("_backfill_cursor")
-        if backfill_remaining > 0 or pending_cursor:
+        in_backfill = overrides.get("_backfill_state") == "running"
+        if in_backfill:
             skip_value: bool | str = BACKFILL_SKIP_VALUE
-            source_config.timeout = BACKFILL_TIMEOUT_SECONDS
+            source_config.timeout = BACKFILL_CHUNK_SECONDS
+            pending_cursor = overrides.get("_backfill_cursor")
             if ctx["platform"] == "patreon" and pending_cursor:
                 source_config.resume_cursor = pending_cursor
         else:
@@ -171,6 +168,24 @@ class DownloadService:
                         skip_value=skip_value,
                     )
 
+        # A backfill chunk that hit its time-box but made forward progress is
+        # NORMAL, not a failure — reclassify TIMEOUT → PARTIAL so it reads as
+        # "ok/progress" (PARTIAL maps to status "ok"), not a red error, and the
+        # next chunk just resumes from the new cursor (plan #693). A chunk that
+        # timed out with NO progress stays TIMEOUT and feeds phase 3's
+        # stall-guard. Leave RATE_LIMITED alone so the platform-cooldown fires.
+        if in_backfill and dl_result.error_type == ErrorType.TIMEOUT:
+            new_cursor = parse_last_cursor(dl_result.stdout, dl_result.stderr)
+            advanced = bool(
+                (new_cursor and new_cursor != overrides.get("_backfill_cursor"))
+                or dl_result.files_downloaded > 0
+            )
+            if advanced:
+                dl_result.error_type = ErrorType.PARTIAL
+                dl_result.error_message = (
+                    f"Backfill chunk: {dl_result.files_downloaded} file(s) — continuing"
+                )
+
         return await self._phase3_persist(
             ctx["event_id"], ctx, dl_result, resolved_campaign_id,
         )
@@ -411,35 +426,38 @@ class DownloadService:
         return event_id
 
     async def _apply_backfill_lifecycle(self, ctx: dict, dl_result) -> None:
-        """Cursor-paged backfill state machine (plan #689).
+        """Backfill state machine (plan #693, building on the cursor of #689).
 
-        A backfill is "in progress" while backfill_runs_remaining > 0 OR a
-        checkpoint cursor is stored. Completion is a CLEAN rc=0 finish —
-        gallery-dl with skip:True exits 0 only after exhausting the
-        newest→oldest walk; a run cut short by the subprocess budget returns
-        success=False / return_code=-1 (the TimeoutExpired path), so rc=0 is
-        an unambiguous "reached the bottom" signal regardless of how many
-        files this run fetched. (This replaces plan #544's files==0 predicate,
-        whose audit-2026-06-02 VALIDATION_FAILED guard still holds: those exit
-        non-zero or carry an error_type, so they are not "completed".)
+        A backfill runs in time-boxed chunks while
+        `config_overrides["_backfill_state"] == "running"`. Each chunk:
+          - COMPLETES the walk → clean rc=0 (gallery-dl with skip:True exits 0
+            only after exhausting the newest→oldest walk; a chunk cut short by
+            its time-box returns success=False / rc<0 via TimeoutExpired). On
+            completion: state="complete", clear the cursor, return to tick mode.
+          - made PROGRESS (cursor advanced and/or files written) → stay
+            "running", checkpoint the new cursor, bump the chunk counter, and
+            spend one of the safety-cap chunks (backfill_runs_remaining). If the
+            cap is exhausted without finishing → state="stalled".
+          - made NO progress → increment the stall counter; two strikes →
+            state="stalled", clear the cursor (a wedged walk can't loop).
 
-        For patreon, each non-completing run checkpoints gallery-dl's
-        last-emitted pagination cursor — even a timed-out run, since its
-        partial stdout/stderr still carries it — so the next run RESUMES
-        instead of re-walking from the top. A stuck-guard clears the cursor
-        after two consecutive runs that fail to advance it, so a wedged walk
-        can't re-strand forever (cf. download soft-limit ladder).
+        State is the single source of truth; the cursor lives only inside a
+        running backfill. config_overrides is reassigned (not mutated in place)
+        for SQLAlchemy JSON change detection.
         """
         overrides = ctx["config_overrides"] or {}
+        if overrides.get("_backfill_state") != "running":
+            return  # not backfilling — tick mode, nothing to do
+
         old_cursor = overrides.get("_backfill_cursor")
-        backfill_remaining = ctx.get("backfill_runs_remaining", 0) or 0
-        if backfill_remaining <= 0 and not old_cursor:
-            return  # tick mode — counter untouched
+        cap_remaining = ctx.get("backfill_runs_remaining", 0) or 0
 
         src = (await self.async_session.execute(
             select(Source).where(Source.id == ctx["source_id"])
         )).scalar_one()
         new_overrides = dict(src.config_overrides or {})
+        chunks = int(new_overrides.get("_backfill_chunks", 0)) + 1
+        new_overrides["_backfill_chunks"] = chunks
 
         completed = (
             dl_result.success
@@ -447,33 +465,43 @@ class DownloadService:
             and dl_result.return_code == 0
         )
         if completed:
+            new_overrides["_backfill_state"] = "complete"
             new_overrides.pop("_backfill_cursor", None)
             new_overrides.pop("_backfill_cursor_stalls", None)
             src.config_overrides = new_overrides
             src.backfill_runs_remaining = 0
             return
 
-        # Non-completing run. Patreon checkpoints + resumes via cursor;
-        # other platforms have no resumable cursor, so just decrement.
-        if ctx["platform"] == "patreon":
-            new_cursor = parse_last_cursor(dl_result.stdout, dl_result.stderr)
-            if new_cursor and new_cursor != old_cursor:
+        # Did not finish. Patreon checkpoints + resumes via cursor; other
+        # platforms have no resumable cursor (every chunk re-walks from the
+        # top), so they advance only by the download archive growing.
+        new_cursor = (
+            parse_last_cursor(dl_result.stdout, dl_result.stderr)
+            if ctx["platform"] == "patreon" else None
+        )
+        advanced = bool(
+            (new_cursor and new_cursor != old_cursor)
+            or dl_result.files_downloaded > 0
+        )
+        if advanced:
+            if new_cursor:
                 new_overrides["_backfill_cursor"] = new_cursor
+            new_overrides.pop("_backfill_cursor_stalls", None)
+            cap_remaining = max(0, cap_remaining - 1)
+            src.backfill_runs_remaining = cap_remaining
+            if cap_remaining == 0:
+                # Safety cap hit before reaching the bottom — pause, don't loop.
+                new_overrides["_backfill_state"] = "stalled"
+        else:
+            stalls = int(new_overrides.get("_backfill_cursor_stalls", 0)) + 1
+            if stalls >= 2:
+                new_overrides["_backfill_state"] = "stalled"
+                new_overrides.pop("_backfill_cursor", None)
                 new_overrides.pop("_backfill_cursor_stalls", None)
             else:
-                stalls = int(new_overrides.get("_backfill_cursor_stalls", 0)) + 1
-                if stalls >= 2:
-                    # Wedged: give up, drop back to tick mode. The surfaced
-                    # error_type on the event tells the operator why.
-                    new_overrides.pop("_backfill_cursor", None)
-                    new_overrides.pop("_backfill_cursor_stalls", None)
-                    src.config_overrides = new_overrides
-                    src.backfill_runs_remaining = 0
-                    return
                 new_overrides["_backfill_cursor_stalls"] = stalls
-            src.config_overrides = new_overrides
 
-        src.backfill_runs_remaining = max(0, backfill_remaining - 1)
+        src.config_overrides = new_overrides
 
     async def _update_source_health(
         self, *, source_id: int, status: str, error_message: str | None,

backend/app/services/extension_service.py

@@ -16,7 +16,7 @@ from sqlalchemy.ext.asyncio import AsyncSession
 
 from ..models import Artist, Source
 from ..utils.slug import slugify
-from .source_service import NEW_SOURCE_BACKFILL_RUNS
+from .source_service import BACKFILL_MAX_CHUNKS
 
 
 class UnknownPlatformError(Exception):
@@ -205,16 +205,17 @@ class ExtensionService:
             return existing, False
         sp = await self.session.begin_nested()
         try:
-            # New subscription sources arm a few backfill runs so the
-            # first ticks walk the full history (otherwise gallery-dl's
-            # exit:20 short-circuits before the archive is built).
-            # Mirrors SourceService.create — without it, Firefox quick-
-            # add on a creator with >20 unsynced posts would surface
-            # as "check failed" with no diagnosis. Audit 2026-06-02.
+            # New subscription sources arm run-until-done backfill (plan #693)
+            # so the first ticks walk the full history (otherwise gallery-dl's
+            # exit:20 short-circuits before the archive is built). Mirrors
+            # SourceService.create — without it, Firefox quick-add on a creator
+            # with >20 unsynced posts would surface as "check failed" with no
+            # diagnosis. Audit 2026-06-02.
             src = Source(
                 artist_id=artist_id, platform=platform,
                 url=url, enabled=True,
-                backfill_runs_remaining=NEW_SOURCE_BACKFILL_RUNS,
+                config_overrides={"_backfill_state": "running"},
+                backfill_runs_remaining=BACKFILL_MAX_CHUNKS,
             )
             self.session.add(src)
             await self.session.flush()

backend/app/services/gallery_dl.py

@@ -56,25 +56,24 @@ class ErrorType(StrEnum):
 # 20 contiguous HEADs is still negligible.
 TICK_SKIP_VALUE = "exit:20"
 
-# Backfill mode (operator-triggered deep scan): walk the full history.
-# Source.backfill_runs_remaining > 0 selects this mode; the longer
-# timeout below absorbs creators with thousands of posts.
+# Backfill mode (operator-triggered deep scan): walk the full history,
+# one TIME-BOXED CHUNK per run (plan #693). config_overrides["_backfill_state"]
+# == "running" selects this mode; the cursor checkpoint (plan #689) lets each
+# chunk resume where the last stopped, so the walk advances across chunks
+# until gallery-dl exits cleanly (= reached the bottom).
 #
-# Sits below download_source's Celery soft_time_limit
-# (DOWNLOAD_SOFT_TIME_LIMIT=1350, tasks/download.py) with ~180s of
-# headroom for phase-3 persist. subprocess.run MUST raise TimeoutExpired
-# before Celery raises SoftTimeLimitExceeded — that exception path
-# captures partial stdout/stderr and finalizes the event; the soft-limit
-# path (until the 2026-06-03 fix) did not. Audit history: 1800 guaranteed
-# SIGKILL against the old hard limit (Knuxy #38275); 1170 was then sized
-# "30s shy of the hard limit (1200)" but still EXCEEDED the soft limit
-# (900), so SoftTimeLimitExceeded preempted TimeoutExpired and every
-# backfill stranded empty (Anduo #39912). Raising the Celery soft/hard
-# limits to 1350/1500 (tasks/download.py) is what made 1170 safe.
-# backfill_runs_remaining=3 still gives ~58 minutes of cumulative walk
-# across three runs for prolific creators.
+# The chunk budget is deliberately FAR below download_source's Celery
+# soft_time_limit (DOWNLOAD_SOFT_TIME_LIMIT=1350, tasks/download.py), not
+# "just under" it. Hitting this budget is the NORMAL chunk boundary, not a
+# failure: subprocess.run raises TimeoutExpired (which captures partial
+# stdout/stderr + the last emitted cursor), and download_service reclassifies
+# a chunk that made progress as PARTIAL (status "ok"), not an error. The huge
+# headroom means a stuck file or a slow chunk can never let Celery's
+# SoftTimeLimitExceeded preempt TimeoutExpired (the failure mode behind Knuxy
+# #38275 / Anduo #39912/#40411). Earlier we ran one ~1170s run-to-the-wall
+# per arming; that died as a timeout error every time on large catalogs.
 BACKFILL_SKIP_VALUE = True
-BACKFILL_TIMEOUT_SECONDS = 1170
+BACKFILL_CHUNK_SECONDS = 600
 
 
 # Sits well below download_source's Celery soft_time_limit
@@ -281,7 +280,12 @@ class GalleryDLService:
                 "skip": True,
                 "sleep": self._rate_limit,
                 "sleep-request": max(0.5, self._rate_limit / 4),
-                "retries": 3,
+                # 2 (not 3) retries — a stuck CDN host shouldn't burn a whole
+                # backfill chunk on one file. Anduo #40838 spent ~600s on a
+                # single image (4 extractor HEAD retries × 30s + 4 downloader
+                # GET retries × 120s); halving retries + the downloader
+                # timeout below caps a wedged file at ~1-2 min instead.
+                "retries": 2,
                 "timeout": 30.0,
                 "verify": True,
                 "postprocessors": [
@@ -296,8 +300,12 @@ class GalleryDLService:
             "downloader": {
                 "part": True,
                 "part-directory": str(self._config_dir / "temp"),
-                "retries": 3,
-                "timeout": 120.0,
+                # See the extractor retries note above (Anduo #40838): 2
+                # retries + a 60s read-timeout (was 120) so a stalled
+                # connection fails fast. 60s is a per-read timeout, not a
+                # transfer cap — large GIFs that keep streaming are unaffected.
+                "retries": 2,
+                "timeout": 60.0,
                 # Forward Patreon as Referer/Origin to yt-dlp when it
                 # fetches video manifests. Operator-flagged 2026-06-01
                 # (DaferQ patreon): video posts hosted on Mux carry a JWT

backend/app/services/source_service.py

@@ -64,6 +64,9 @@ class SourceRecord:
     consecutive_failures: int
     next_check_at: str | None
     backfill_runs_remaining: int
+    # plan #693: derived from config_overrides for the UI badge.
+    backfill_state: str | None  # "running" | "complete" | "stalled" | None (idle)
+    backfill_chunks: int
 
     def to_dict(self) -> dict:
         return {
@@ -82,6 +85,8 @@ class SourceRecord:
             "consecutive_failures": self.consecutive_failures,
             "next_check_at": self.next_check_at,
             "backfill_runs_remaining": self.backfill_runs_remaining,
+            "backfill_state": self.backfill_state,
+            "backfill_chunks": self.backfill_chunks,
         }
 
 
@@ -89,10 +94,13 @@ class SourceRecord:
 
 _EDITABLE = {"enabled", "url", "config_overrides", "check_interval_override", "platform"}
 
-# Plan #544 follow-up: newly created enabled sources pre-arm backfill so
-# their first N polls walk gallery-dl's full post history with the longer
-# timeout (matches the manual "Deep scan" button's default).
-NEW_SOURCE_BACKFILL_RUNS = 3
+# Plan #693: backfill safety cap. "Start backfill" (and a newly created
+# enabled source) arms a run-until-done walk; this caps how many time-boxed
+# chunks it may spend before pausing as "stalled", so a pathological walk that
+# never reaches the bottom can't run forever. Generous on purpose — at
+# BACKFILL_CHUNK_SECONDS (600s) per chunk this is ~33h of cumulative walk, far
+# beyond any real catalog; the cursor stall-guard is the real terminator.
+BACKFILL_MAX_CHUNKS = 200
 
 
 class SourceService:
@@ -135,6 +143,7 @@ class SourceService:
         self, source: Source, artist: Artist, settings: ImportSettings,
     ) -> SourceRecord:
         nxt = compute_next_check_at(source, artist, settings)
+        co = source.config_overrides or {}
         return SourceRecord(
             id=source.id,
             artist_id=source.artist_id,
@@ -151,6 +160,8 @@ class SourceService:
             consecutive_failures=source.consecutive_failures or 0,
             next_check_at=nxt.isoformat() if nxt else None,
             backfill_runs_remaining=source.backfill_runs_remaining or 0,
+            backfill_state=co.get("_backfill_state"),
+            backfill_chunks=int(co.get("_backfill_chunks", 0)),
         )
 
     async def _row_to_record(self, source: Source) -> SourceRecord:
@@ -212,16 +223,17 @@ class SourceService:
             select(func.count(Source.id)).where(Source.artist_id == artist_id)
         )).scalar_one()
 
-        # Plan #544 follow-up: a freshly added subscription has no archive
-        # yet, so the first few polls would walk the full post history in
-        # tick mode and trip exit:20 after ~20 contiguous archive hits —
-        # except there are none yet, so tick mode would walk forever and
-        # blow the wall-clock cap. Pre-arm backfill so the initial syncs
-        # use the longer timeout + skip:True walk. Tick mode resumes once
-        # the budget is spent or the queue drains.
-        # Disabled sources (incl. sidecar synthetics, url='sidecar:...')
-        # are never polled, so leave their counter at 0.
-        backfill_runs = NEW_SOURCE_BACKFILL_RUNS if enabled else 0
+        # Plan #693: a freshly added subscription has no archive yet, so it
+        # should walk its full post history once. Arm run-until-done backfill
+        # (state="running" + the chunk cap); the time-boxed chunks march to the
+        # bottom across ticks, then flip to "complete" and tick mode takes over.
+        # Disabled sources (incl. sidecar synthetics, url='sidecar:...') are
+        # never polled, so leave them idle.
+        if enabled:
+            config_overrides = {**(config_overrides or {}), "_backfill_state": "running"}
+            backfill_runs = BACKFILL_MAX_CHUNKS
+        else:
+            backfill_runs = 0
         source = Source(
             artist_id=artist_id, platform=platform, url=url,
             enabled=enabled, config_overrides=config_overrides,
@@ -286,22 +298,41 @@ class SourceService:
         await self.session.commit()
         return await self._row_to_record(source)
 
-    async def set_backfill_runs(
-        self, source_id: int, runs: int,
-    ) -> SourceRecord:
-        """Plan #544: arm a source for backfill mode. The next `runs`
-        download runs will use gallery-dl's full-walk config (skip: True
-        + 30-min timeout) instead of the catch-up default. Runs must be
-        1..10 — bigger is rejected to keep the operator from accidentally
-        setting a runaway budget."""
-        if not isinstance(runs, int) or runs < 1 or runs > 10:
-            raise ValueError("runs must be an integer in [1, 10]")
+    async def start_backfill(self, source_id: int) -> SourceRecord:
+        """Plan #693: arm a run-until-done backfill. Sets state="running" and
+        the chunk cap; download runs then walk the full post history in
+        time-boxed chunks (skip:True + BACKFILL_CHUNK_SECONDS), resuming from
+        the cursor each chunk, until gallery-dl reaches the bottom (→ state
+        "complete") or the cap/stall-guard pauses it (→ "stalled"). Clears any
+        prior cursor/chunk/stall state so a re-start walks fresh from the top."""
         source = (await self.session.execute(
             select(Source).where(Source.id == source_id)
         )).scalar_one_or_none()
         if source is None:
             raise LookupError(f"source id={source_id} not found")
-        source.backfill_runs_remaining = runs
+        co = dict(source.config_overrides or {})
+        co["_backfill_state"] = "running"
+        for k in ("_backfill_cursor", "_backfill_cursor_stalls", "_backfill_chunks"):
+            co.pop(k, None)
+        source.config_overrides = co
+        source.backfill_runs_remaining = BACKFILL_MAX_CHUNKS
+        await self.session.commit()
+        return await self._row_to_record(source)
+
+    async def stop_backfill(self, source_id: int) -> SourceRecord:
+        """Plan #693: cancel an in-progress backfill — back to idle/tick mode.
+        Clears the running state + cursor/chunk/stall bookkeeping."""
+        source = (await self.session.execute(
+            select(Source).where(Source.id == source_id)
+        )).scalar_one_or_none()
+        if source is None:
+            raise LookupError(f"source id={source_id} not found")
+        co = dict(source.config_overrides or {})
+        for k in ("_backfill_state", "_backfill_cursor", "_backfill_cursor_stalls",
+                  "_backfill_chunks"):
+            co.pop(k, None)
+        source.config_overrides = co
+        source.backfill_runs_remaining = 0
         await self.session.commit()
         return await self._row_to_record(source)
 

backend/app/tasks/download.py

@@ -31,7 +31,7 @@ _KEY_PATH = IMAGES_ROOT / "secrets" / "credential_key.b64"
 # SoftTimeLimitExceeded in-process, whereas the HARD limit SIGKILLs the
 # worker (no chance to finalize). Both gallery-dl subprocess budgets
 # (gallery_dl.py: _DEFAULT_GDL_TIMEOUT_SECONDS=870 tick,
-# BACKFILL_TIMEOUT_SECONDS=1170 backfill) MUST sit below the soft limit
+# BACKFILL_CHUNK_SECONDS=600 per backfill chunk, plan #693) MUST sit below the soft limit
 # so subprocess.run raises its own TimeoutExpired first — that path
 # captures partial stdout/stderr and finalizes the DownloadEvent. soft is
 # max-subprocess (1170) + ~180s phase-3 persist headroom; hard is soft +