diff --git a/docs/superpowers/plans/2026-03-18-comprehensive-review.md b/docs/superpowers/plans/2026-03-18-comprehensive-review.md
new file mode 100644
index 0000000..e5afc78
--- /dev/null
+++ b/docs/superpowers/plans/2026-03-18-comprehensive-review.md
@@ -0,0 +1,862 @@
+# GallerySubscriber Comprehensive Review Implementation Plan
+
+> **For agentic workers:** REQUIRED SUB-SKILL: Use superpowers:subagent-driven-development (recommended) or superpowers:executing-plans to implement this plan task-by-task. Steps use checkbox (`- [ ]`) syntax for tracking.
+
+**Goal:** Fix four confirmed bugs, replace the query-time superseded subquery with a write-time column, improve error classification, and clean up code quality issues across the backend and frontend.
+
+**Architecture:** Layered delivery — bugs first (safe, atomic), then the superseded DB migration + logic, then error classification restructuring, then quality cleanup. Each task produces a working commit.
+
+**Tech Stack:** Python 3.10+, Quart, SQLAlchemy 2 async, Alembic, Celery, Vue 3, PostgreSQL 15
+
+**Spec:** `docs/superpowers/specs/2026-03-18-comprehensive-review-design.md`
+
+**Note on tests:** This project has no existing test infrastructure. Steps include manual verification commands instead of automated tests.
+
+---
+
+## Files Modified
+
+| File | Change |
+|------|--------|
+| `backend/app/tasks/downloads.py` | Bug 1 fix (pass download_id); Section 2 write logic |
+| `backend/app/tasks/maintenance.py` | Bug 2 fix (started_at message); import from db.py |
+| `backend/app/services/gallery_dl.py` | Bug 3 (get_running_loop); Bug 4 (utcnow); Sec 3a/3b/3c; 4c |
+| `backend/app/models/download.py` | Add `superseded` column + `to_dict()` entry |
+| `backend/alembic/versions/003_add_superseded.py` | New migration: column + backfill + index |
+| `backend/app/api/downloads.py` | Remove `_not_superseded_condition()`; replace with `superseded==False`; fix sessions |
+| `backend/app/api/credentials.py` | Fix unclosed sessions |
+| `backend/app/api/settings.py` | Fix unclosed sessions |
+| `backend/app/api/sources.py` | Fix unclosed sessions |
+| `backend/app/api/subscriptions.py` | Fix unclosed sessions |
+| `frontend/src/views/Dashboard.vue` | Parallelize `checkAllSources` |
+| `backend/app/tasks/db.py` | **New file** — shared `get_async_session` + `_cleanup_engine` |
+
+---
+
+## Task 1: Fix Bug 1 — process_download doesn't pass download_id
+
+**Files:**
+- Modify: `backend/app/tasks/downloads.py:574`
+
+- [ ] **Step 1: Open the file and locate the `_process` inner function inside `process_download`**
+
+  In `backend/app/tasks/downloads.py`, find the `process_download` task (around line 551). Inside it is an `async def _process():` function. Near the end of `_process`, around line 574, is:
+  ```python
+  return await _download_source_async(download.source_id)
+  ```
+
+- [ ] **Step 2: Fix the call to pass download_id**
+
+  Change that line to:
+  ```python
+  return await _download_source_async(download.source_id, download_id=download_id)
+  ```
+  The `download_id` variable is the parameter of the outer `process_download(self, download_id: int)` function and is in scope for the inner `_process` closure.
+
+- [ ] **Step 3: Verify the change looks correct**
+
+  The `_process` function body should now end with:
+  ```python
+  if not download.source_id:
+      return {"download_id": download_id, "status": "error", "error": "No source associated"}
+
+  # Delegate to download_source
+  return await _download_source_async(download.source_id, download_id=download_id)
+  ```
+
+- [ ] **Step 4: Commit**
+
+  ```bash
+  cd /home/bvandeusen/Nextcloud/Projects/GallerySubscriber
+  git add backend/app/tasks/downloads.py
+  git commit -m "fix: pass download_id in process_download to prevent retry skip"
+  ```
+
+---
+
+## Task 2: Fix Bug 2 — Orphaned job message always shows None
+
+**Files:**
+- Modify: `backend/app/tasks/maintenance.py:176-178`
+
+- [ ] **Step 1: Locate the loop in `_reset_orphaned_running_jobs_async`**
+
+  Find the `for job in orphaned_jobs:` loop. The current code:
+  ```python
+  for job in orphaned_jobs:
+      job.status = DownloadStatus.QUEUED
+      job.started_at = None
+      job.error_message = f"Reset from orphaned running state (was running since {job.started_at})"
+  ```
+
+- [ ] **Step 2: Capture started_at before clearing it**
+
+  Replace those four lines with:
+  ```python
+  for job in orphaned_jobs:
+      original_started_at = job.started_at
+      job.status = DownloadStatus.QUEUED
+      job.started_at = None
+      job.error_message = f"Reset from orphaned running state (was running since {original_started_at})"
+  ```
+
+- [ ] **Step 3: Commit**
+
+  ```bash
+  git add backend/app/tasks/maintenance.py
+  git commit -m "fix: capture started_at before clearing in orphaned job reset message"
+  ```
+
+---
+
+## Task 3: Fix Bugs 3 and 4 — get_event_loop + datetime.utcnow in gallery_dl.py
+
+**Files:**
+- Modify: `backend/app/services/gallery_dl.py`
+
+- [ ] **Step 1: Fix get_event_loop (Bug 3)**
+
+  Find line 616 (inside the `download` method, in the `try` block):
+  ```python
+  loop = asyncio.get_event_loop()
+  ```
+  Change to:
+  ```python
+  loop = asyncio.get_running_loop()
+  ```
+
+- [ ] **Step 2: Fix datetime.utcnow (Bug 4)**
+
+  At the top of the file, add the import after the existing imports:
+  ```python
+  from app.models.base import utcnow
+  ```
+
+  Then find the two `datetime.utcnow()` calls in the `download` method (around lines 561 and 628):
+  ```python
+  started_at = datetime.utcnow().isoformat()
+  ```
+  and:
+  ```python
+  completed_at = datetime.utcnow().isoformat()
+  ```
+  Change both to:
+  ```python
+  started_at = utcnow().isoformat()
+  ```
+  and:
+  ```python
+  completed_at = utcnow().isoformat()
+  ```
+
+  There is also a `datetime.utcnow().isoformat()` in the `except subprocess.TimeoutExpired` block and the `except Exception` block — change those too. Search for all occurrences:
+  ```bash
+  grep -n "datetime.utcnow" backend/app/services/gallery_dl.py
+  ```
+
+- [ ] **Step 3: Remove the now-dead local datetime import inside the `download` method**
+
+  The `download` method has a local import block near the top of its body (around line 557):
+  ```python
+  import asyncio
+  from datetime import datetime
+  ```
+  After replacing all `datetime.utcnow()` calls with `utcnow()`, the `from datetime import datetime`
+  line is no longer used inside this method. Remove it. The `import asyncio` line can stay (it is
+  used for `asyncio.get_running_loop()`).
+
+- [ ] **Step 4: Verify no remaining datetime.utcnow calls**
+
+  ```bash
+  grep -n "datetime.utcnow" backend/app/services/gallery_dl.py
+  # Expected: no output
+  ```
+
+- [ ] **Step 5: Commit**
+
+  ```bash
+  git add backend/app/services/gallery_dl.py
+  git commit -m "fix: replace get_event_loop with get_running_loop and utcnow consistently"
+  ```
+
+---
+
+## Task 4: Add superseded column to Download model
+
+**Files:**
+- Modify: `backend/app/models/download.py`
+
+- [ ] **Step 1: Add the column import**
+
+  At the top of the file, `Boolean` is not yet imported from sqlalchemy. Add it to the import:
+  ```python
+  from sqlalchemy import String, Integer, Text, ForeignKey, BigInteger, Boolean
+  ```
+
+- [ ] **Step 2: Add the mapped column to the Download class**
+
+  After the `total_size` column (around line 56), add:
+  ```python
+  # Superseded: True if a later successful download has run for this source
+  superseded: Mapped[bool] = mapped_column(Boolean, default=False, nullable=False, server_default="false")
+  ```
+
+- [ ] **Step 3: Add to to_dict()**
+
+  In the `to_dict` method, add `"superseded": self.superseded` after `"completed_at"`:
+  ```python
+  "completed_at": format_datetime(self.completed_at),
+  "superseded": self.superseded,
+  "metadata": self.metadata_,
+  ```
+
+- [ ] **Step 4: Commit**
+
+  ```bash
+  git add backend/app/models/download.py
+  git commit -m "feat: add superseded column to Download model"
+  ```
+
+---
+
+## Task 5: Create Alembic migration for superseded column
+
+**Files:**
+- Create: `backend/alembic/versions/003_add_superseded.py`
+
+- [ ] **Step 1: Create the migration file**
+
+  Create `backend/alembic/versions/003_add_superseded.py` with this content:
+
+  ```python
+  """Add superseded column to downloads
+
+  Revision ID: 003
+  Revises: 002
+  Create Date: 2026-03-18
+
+  """
+  from typing import Sequence, Union
+
+  from alembic import op
+  import sqlalchemy as sa
+
+
+  revision: str = '003'
+  down_revision: Union[str, None] = '002'
+  branch_labels: Union[str, Sequence[str], None] = None
+  depends_on: Union[str, Sequence[str], None] = None
+
+
+  def upgrade() -> None:
+      # Add superseded column with default False
+      op.add_column(
+          'downloads',
+          sa.Column('superseded', sa.Boolean(), nullable=False, server_default='false')
+      )
+
+      # Backfill: mark FAILED downloads as superseded where a later COMPLETED
+      # download exists for the same source. REQUIRED — without this, all
+      # previously-superseded failures reappear in the dashboard immediately.
+      op.execute("""
+          UPDATE downloads d
+          SET superseded = TRUE
+          WHERE d.status = 'failed'
+            AND EXISTS (
+              SELECT 1 FROM downloads d2
+              WHERE d2.source_id = d.source_id
+                AND d2.status = 'completed'
+                AND d2.created_at > d.created_at
+            )
+      """)
+
+      # Partial index covering the dominant query pattern: status + superseded = FALSE.
+      # A single boolean column index would be ignored in favour of status indexes;
+      # this form matches the WHERE clauses at all three affected query sites.
+      op.create_index(
+          'idx_downloads_not_superseded',
+          'downloads',
+          ['status'],
+          postgresql_where=sa.text('superseded = FALSE')
+      )
+
+
+  def downgrade() -> None:
+      op.drop_index('idx_downloads_not_superseded', table_name='downloads')
+      op.drop_column('downloads', 'superseded')
+  ```
+
+- [ ] **Step 2: Run the migration**
+
+  ```bash
+  cd /home/bvandeusen/Nextcloud/Projects/GallerySubscriber
+  docker compose exec app alembic upgrade head
+  ```
+
+  Expected output: `Running upgrade 002 -> 003, Add superseded column to downloads`
+
+- [ ] **Step 3: Verify the column and index exist**
+
+  ```bash
+  docker compose exec db psql -U postgres -d gallerysubscriber -c "\d downloads" | grep superseded
+  # Expected: superseded | boolean | not null | false
+  docker compose exec db psql -U postgres -d gallerysubscriber -c "\di idx_downloads_not_superseded"
+  # Expected: index listed
+  ```
+
+- [ ] **Step 4: Commit**
+
+  ```bash
+  git add backend/alembic/versions/003_add_superseded.py
+  git commit -m "feat: migration to add superseded column with backfill and partial index"
+  ```
+
+---
+
+## Task 6: Replace _not_superseded_condition() with superseded==False in api/downloads.py
+
+**Files:**
+- Modify: `backend/app/api/downloads.py`
+
+- [ ] **Step 1: Remove `exists` from imports if no longer needed**
+
+  After removing `_not_superseded_condition()`, the `exists` import in `api/downloads.py` becomes
+  unused. The current import line is:
+  ```python
+  from sqlalchemy import select, func, and_, or_, desc, exists
+  ```
+  Check whether `exists` is still referenced anywhere in the file:
+  ```bash
+  grep -n "exists" backend/app/api/downloads.py
+  ```
+  If no remaining uses, remove `exists` from the import.
+
+- [ ] **Step 2: Replace the `_not_superseded_condition()` function**
+
+  Delete the entire `_not_superseded_condition()` function (lines 18-29). It will no longer be used.
+
+- [ ] **Step 3: Update `list_downloads` — exclude_superseded filter**
+
+  Find the block around line 63:
+  ```python
+  if exclude_superseded and status == DownloadStatus.FAILED:
+      filters.append(_not_superseded_condition())
+  ```
+  Replace with:
+  ```python
+  if exclude_superseded and status == DownloadStatus.FAILED:
+      filters.append(Download.superseded == False)
+  ```
+
+- [ ] **Step 4: Update `recent_activity` — failed filter**
+
+  Find the `or_` block around line 188:
+  ```python
+  and_(
+      Download.status == DownloadStatus.FAILED,
+      _not_superseded_condition(),
+  ),
+  ```
+  Replace with:
+  ```python
+  and_(
+      Download.status == DownloadStatus.FAILED,
+      Download.superseded == False,
+  ),
+  ```
+
+- [ ] **Step 5: Update `get_stats` — active_failed query**
+
+  Find the `active_failed_query` block around line 331:
+  ```python
+  active_failed_query = select(func.count()).select_from(Download).where(
+      and_(
+          Download.status == DownloadStatus.FAILED,
+          _not_superseded_condition(),
+      )
+  )
+  ```
+  Replace with:
+  ```python
+  active_failed_query = select(func.count()).select_from(Download).where(
+      and_(
+          Download.status == DownloadStatus.FAILED,
+          Download.superseded == False,
+      )
+  )
+  ```
+
+- [ ] **Step 6: Remove the unused `exists` import if no longer needed**
+
+  Check: `grep -n "exists" backend/app/api/downloads.py`
+  If `exists` is no longer referenced, remove it from the sqlalchemy import line.
+
+- [ ] **Step 7: Verify no remaining references to _not_superseded_condition**
+
+  ```bash
+  grep -rn "_not_superseded_condition" backend/
+  # Expected: no output
+  ```
+
+- [ ] **Step 8: Commit**
+
+  ```bash
+  git add backend/app/api/downloads.py
+  git commit -m "feat: replace _not_superseded_condition subquery with superseded column filter"
+  ```
+
+---
+
+## Task 7: Add superseded write logic to download task
+
+**Files:**
+- Modify: `backend/app/tasks/downloads.py`
+
+- [ ] **Step 1: Add `update` to imports**
+
+  The current sqlalchemy import line in `backend/app/tasks/downloads.py` is:
+  ```python
+  from sqlalchemy import select, and_, or_
+  ```
+  Change it to:
+  ```python
+  from sqlalchemy import select, and_, or_, update
+  ```
+
+- [ ] **Step 2: Locate the COMPLETED path in Phase 3**
+
+  In `_download_source_async`, find Phase 3 (around line 297). Find the `if dl_result.success:` block:
+  ```python
+  if dl_result.success:
+      download.status = DownloadStatus.COMPLETED
+      source.last_success = utcnow()
+      source.error_count = 0
+      logger.info(...)
+
+      publish_event("download.completed", { ... })
+  ```
+
+- [ ] **Step 3: Insert the bulk supersede UPDATE before publish_event**
+
+  Insert the UPDATE after `source.error_count = 0` and before `publish_event`:
+  ```python
+  if dl_result.success:
+      download.status = DownloadStatus.COMPLETED
+      source.last_success = utcnow()
+      source.error_count = 0
+      logger.info(f"Download completed for {subscription_name}/{source_platform}: {dl_result.files_downloaded} files")
+
+      # Mark any prior failures for this source as superseded
+      await session.execute(
+          update(Download)
+          .where(
+              Download.source_id == source_id,
+              Download.status == DownloadStatus.FAILED,
+          )
+          .values(superseded=True)
+      )
+
+      publish_event("download.completed", {
+          "download_id": actual_download_id,
+          "source_id": source_id,
+          "subscription_name": subscription_name,
+          "platform": source_platform,
+          "file_count": dl_result.files_downloaded,
+          "duration_seconds": dl_result.duration_seconds,
+      })
+  ```
+
+- [ ] **Step 4: Verify the ordering is correct**
+
+  The `if dl_result.success:` block should contain (in order):
+  1. Set `download.status = COMPLETED`
+  2. Update source success fields (`last_success`, `error_count`)
+  3. Log the completion
+  4. Bulk UPDATE prior failures to superseded
+  5. `publish_event("download.completed", ...)`
+
+  **Outside** the `if/else` block (shared by both success and failure paths, unchanged):
+  - `source.last_check = utcnow()`
+  - `await session.commit()`
+
+  Do not move `source.last_check` or `session.commit()` inside the success block — they currently
+  sit after the entire `if dl_result.success: ... else: ...` structure and must remain there.
+
+- [ ] **Step 5: Commit**
+
+  ```bash
+  git add backend/app/tasks/downloads.py
+  git commit -m "feat: mark prior failures as superseded on successful download"
+  ```
+
+---
+
+## Task 8: Fix unclosed AsyncSession in all API blueprints
+
+The pattern `session = AsyncSession(bind=conn)` is used 31 times across 5 API files. All need changing to `async with AsyncSession(bind=conn) as session:` with the body indented inside the `with` block.
+
+**Files:**
+- Modify: `backend/app/api/downloads.py`
+- Modify: `backend/app/api/credentials.py`
+- Modify: `backend/app/api/settings.py`
+- Modify: `backend/app/api/sources.py`
+- Modify: `backend/app/api/subscriptions.py`
+
+The pattern to apply in each handler:
+
+**Before:**
+```python
+async with current_app.db_engine.connect() as conn:
+    session = AsyncSession(bind=conn)
+    # ... session usage ...
+    return jsonify(...)
+```
+
+**After:**
+```python
+async with current_app.db_engine.connect() as conn:
+    async with AsyncSession(bind=conn) as session:
+        # ... session usage (indented one level deeper) ...
+        return jsonify(...)
+```
+
+- [ ] **Step 1: Fix downloads.py**
+
+  Apply the pattern to all 6 handlers in `backend/app/api/downloads.py`. Each `async with current_app.db_engine.connect() as conn:` block gets an inner `async with AsyncSession(bind=conn) as session:` with the session body re-indented.
+
+- [ ] **Step 2: Fix credentials.py**
+
+  Apply to all 5 occurrences in `backend/app/api/credentials.py`.
+
+- [ ] **Step 3: Fix settings.py**
+
+  Apply to all 6 occurrences in `backend/app/api/settings.py`.
+
+- [ ] **Step 4: Fix sources.py**
+
+  Apply to all 6 occurrences in `backend/app/api/sources.py`.
+
+- [ ] **Step 5: Fix subscriptions.py**
+
+  Apply to all 8 occurrences in `backend/app/api/subscriptions.py`.
+
+- [ ] **Step 6: Verify no bare session assignments remain**
+
+  ```bash
+  grep -rn "session = AsyncSession(bind=" backend/app/api/
+  # Expected: no output
+  ```
+
+- [ ] **Step 7: Commit**
+
+  ```bash
+  git add backend/app/api/
+  git commit -m "fix: wrap AsyncSession in context manager across all API blueprints"
+  ```
+
+---
+
+## Task 9: Error classification — extract patterns, add logging, fix return code 1
+
+**Files:**
+- Modify: `backend/app/services/gallery_dl.py`
+
+- [ ] **Step 1: Extract pattern lists to class-level constants**
+
+  Inside the `GalleryDLService` class, before `PLATFORM_DEFAULTS`, add these class-level constants:
+
+  ```python
+  # Error classification patterns — used by _categorize_error()
+  AUTH_ERROR_PATTERNS = [
+      "unauthorized",
+      "login required",
+      "please login",
+      "must be logged in",
+      "authentication required",
+      "authenticationerror",
+      "refresh-token",
+      "session expired",
+      "invalid cookie",
+      "cookies are expired",
+      "cookies have expired",
+      "not logged in",
+  ]
+  RATE_LIMIT_PATTERNS = [
+      '" 429 ',
+      "rate limit",
+      "too many requests",
+      "ratelimit",
+      "throttl",
+  ]
+  NOT_FOUND_PATTERNS = [
+      '" 404 ',
+      "error 404",
+      "status: 404",
+      "status code: 404",
+      "404 not found",
+      "page not found",
+      "user not found",
+      "creator not found",
+      "artist not found",
+      "content no longer available",
+      "has been deleted",
+      "account.*deleted",
+      "profile.*not.*exist",
+  ]
+  GENERIC_NOT_FOUND_PATTERNS = ["does not exist", "no longer available"]
+  NETWORK_ERROR_PATTERNS = [
+      "timed out",
+      "read timed out",
+      "connect timed out",
+      "connection timed out",
+      "connection refused",
+      "connection reset",
+      "network unreachable",
+      "name resolution failed",
+      "name or service not known",
+      "nodename nor servname provided",
+      "ssl: certificate_verify_failed",
+      "ssl: wrong_version_number",
+      "certificate verify failed",
+      "[errno",
+  ]
+  ACCESS_DENIED_PATTERNS = [
+      '" 403 ',
+      "error 403",
+      "forbidden",
+      "access denied",
+      "permission denied",
+      "tier required",
+      "pledge required",
+  ]
+  ```
+
+- [ ] **Step 2: Rewrite `_categorize_error` to use the constants and add debug logging**
+
+  Replace the inline lists in `_categorize_error` with references to the class constants (e.g. `self.AUTH_ERROR_PATTERNS`). Wrap each `return` statement with a debug log. Example for auth errors:
+
+  ```python
+  has_401_error = (
+      '" 401 ' in combined or
+      "401 unauthorized" in combined or
+      "error 401" in combined or
+      "status: 401" in combined or
+      "status code: 401" in combined
+  )
+  if has_401_error or any(pattern in combined for pattern in self.AUTH_ERROR_PATTERNS):
+      matched = "401 HTTP response" if has_401_error else next(p for p in self.AUTH_ERROR_PATTERNS if p in combined)
+      logger.debug(f"Error classified as auth_error via pattern: '{matched}'")
+      return ErrorType.AUTH_ERROR, "Authentication failed - cookies may be expired or invalid"
+  ```
+
+  Apply the same pattern for all other error types. For `NO_NEW_CONTENT` returns, log the trigger (skip line count, skip text, or empty stdout).
+
+- [ ] **Step 3: Fix return code 1 empty stdout (3c)**
+
+  Find the empty-stdout check in Phase 1 (after the skip detection):
+  ```python
+  if return_code in (0, 1) and not stdout.strip():
+      return ErrorType.NO_NEW_CONTENT, "No new content to download"
+  ```
+  Change to:
+  ```python
+  if return_code == 0 and not stdout.strip():
+      logger.debug("Error classified as no_new_content via: empty stdout with return code 0")
+      return ErrorType.NO_NEW_CONTENT, "No new content to download"
+  ```
+
+- [ ] **Step 4: Commit**
+
+  ```bash
+  git add backend/app/services/gallery_dl.py
+  git commit -m "refactor: extract error patterns to constants, add debug logging, fix rc1 no_new_content"
+  ```
+
+---
+
+## Task 10: Fix _count_downloaded_files heuristic
+
+**Files:**
+- Modify: `backend/app/services/gallery_dl.py`
+
+- [ ] **Step 1: Locate the method**
+
+  Find `_count_downloaded_files` (around line 517). Current implementation:
+  ```python
+  for line in stdout.strip().split("\n"):
+      line = line.strip()
+      if line and not line.startswith("#") and "skip" not in line.lower():
+          count += 1
+  ```
+
+- [ ] **Step 2: Replace with absolute path heuristic**
+
+  With `--verbose` enabled, gallery-dl writes downloaded file paths to stdout as absolute paths (starting with `/`). Log lines use `[` as a prefix character. HTTP lines go to stderr. Count only lines starting with `/`:
+
+  ```python
+  for line in stdout.strip().split("\n"):
+      line = line.strip()
+      if line.startswith("/"):
+          count += 1
+  ```
+
+- [ ] **Step 3: Commit**
+
+  ```bash
+  git add backend/app/services/gallery_dl.py
+  git commit -m "fix: count only absolute path lines in _count_downloaded_files"
+  ```
+
+---
+
+## Task 11: Extract shared DB helpers to tasks/db.py
+
+**Files:**
+- Create: `backend/app/tasks/db.py`
+- Modify: `backend/app/tasks/downloads.py`
+- Modify: `backend/app/tasks/maintenance.py`
+
+- [ ] **Step 1: Create backend/app/tasks/db.py**
+
+  ```python
+  """Shared async database helpers for Celery tasks.
+
+  Provides get_async_session() and _cleanup_engine() used by both
+  downloads.py and maintenance.py. This module must not import from
+  either of those files to avoid circular imports.
+  """
+
+  from sqlalchemy.ext.asyncio import create_async_engine, AsyncSession, async_sessionmaker
+
+  from app.config import get_settings
+
+
+  def get_async_session():
+      """Get async session factory for Celery tasks.
+
+      Creates a fresh engine and session factory each time to avoid
+      event loop issues when asyncio.run() creates new loops.
+      """
+      settings = get_settings()
+      engine = create_async_engine(settings.async_database_url, echo=False)
+      session_factory = async_sessionmaker(engine, class_=AsyncSession, expire_on_commit=False)
+      return session_factory, engine
+
+
+  async def cleanup_engine(engine):
+      """Properly dispose of the async engine."""
+      await engine.dispose()
+  ```
+
+- [ ] **Step 2: Update downloads.py to import from db.py**
+
+  At the top of `backend/app/tasks/downloads.py`, remove the local definitions of `get_async_session` and `_cleanup_engine` (the two functions defined around lines 29-42).
+
+  Add this import:
+  ```python
+  from app.tasks.db import get_async_session, cleanup_engine as _cleanup_engine
+  ```
+
+- [ ] **Step 3: Update maintenance.py to import from db.py**
+
+  At the top of `backend/app/tasks/maintenance.py`, remove the local definitions of `get_async_session` and `_cleanup_engine` (around lines 32-41).
+
+  Add this import:
+  ```python
+  from app.tasks.db import get_async_session, cleanup_engine as _cleanup_engine
+  ```
+
+- [ ] **Step 4: Verify no duplicate definitions remain**
+
+  ```bash
+  grep -n "def get_async_session\|def _cleanup_engine\|def cleanup_engine" \
+    backend/app/tasks/downloads.py \
+    backend/app/tasks/maintenance.py \
+    backend/app/tasks/db.py
+  # Expected: only db.py shows the definitions
+  ```
+
+- [ ] **Step 5: Commit**
+
+  ```bash
+  git add backend/app/tasks/db.py backend/app/tasks/downloads.py backend/app/tasks/maintenance.py
+  git commit -m "refactor: extract shared DB helpers to tasks/db.py"
+  ```
+
+---
+
+## Task 12: Parallelize checkAllSources in Dashboard
+
+**Files:**
+- Modify: `frontend/src/views/Dashboard.vue`
+
+- [ ] **Step 1: Locate checkAllSources**
+
+  Find the `async function checkAllSources()` (around line 460). Current implementation uses a `for...of` loop with `await` inside.
+
+- [ ] **Step 2: Replace with Promise.allSettled**
+
+  Replace the body of `checkAllSources` with:
+
+  ```javascript
+  async function checkAllSources() {
+    checkingAll.value = true
+    try {
+      const enabledSources = sourcesStore.sources.filter(s => s.enabled)
+      const results = await Promise.allSettled(
+        enabledSources.map(source => sourcesStore.triggerCheck(source.id))
+      )
+      const queued = results.filter(r => r.status === 'fulfilled').length
+      const failed = results.filter(r => r.status === 'rejected').length
+      if (failed > 0) {
+        notifications.success(`Queued ${queued} sources (${failed} failed to queue)`)
+      } else {
+        notifications.success(`Queued checks for ${queued} sources`)
+      }
+      await fetchActiveDownloads()
+      await downloadsStore.fetchStats()
+    } catch (error) {
+      notifications.error(`Failed to check sources: ${error.message}`)
+    } finally {
+      checkingAll.value = false
+    }
+  }
+  ```
+
+- [ ] **Step 3: Commit**
+
+  ```bash
+  git add frontend/src/views/Dashboard.vue
+  git commit -m "perf: parallelize checkAllSources with Promise.allSettled"
+  ```
+
+---
+
+## Final Verification
+
+- [ ] **Start the stack and verify no startup errors**
+
+  ```bash
+  docker compose up -d --build
+  docker compose logs -f app scheduler worker
+  ```
+
+- [ ] **Run the migration if not already done**
+
+  ```bash
+  docker compose exec app alembic upgrade head
+  ```
+
+- [ ] **Smoke test the dashboard**
+
+  1. Open the app, confirm the dashboard loads without errors
+  2. Confirm "Sources Needing Attention" shows only active (unsuperseded) failures
+  3. Trigger a retry on a failed download — confirm it actually runs (no longer silently skips)
+  4. After a successful download completes, confirm prior failures for that source disappear from dashboard
+
+- [ ] **Verify the export-failed-logs endpoint returns `superseded` field**
+
+  ```bash
+  curl -s "http://localhost:8080/api/downloads?status=failed&per_page=1" | python3 -m json.tool | grep superseded
+  # Expected: "superseded": false
+  ```