Clean up test index after all tests.

Various things sould return a 404: * invalid primary key in the URL * token doesn't start with 'subscribe' * subscri

activate_archive should pass allow_skip_resync=True to SchedulePremiumSetup so that sync_redis is skipped for large feeds during the concurren

activate_archive should trigger the archive feed fetch.

A late-arriving customer.subscription.created event for the original premium plan must not strip archive/pro from a user who already u

activate_premium (non-archive) should NOT pass allow_skip_resync, because there is no concurrent archive fetch to conflict with.

Stripe sometimes fires the same subscription event twice. The second activate_premium call must be a no-op: no email, no flag flips.

Creating a folder with different case should not create a duplicate.

POST to add_url without 'url' should return error, not crash.

POST to add_url should reject private/link-local addresses before fetching.

The NYT/Techmeme case that motivated this fix.

The /reader/all_classifiers endpoint should put global classifiers in scoped_classifiers and feed-level classifiers in folders — neve

Regular article URLs should not be blocked.

Tokens mixing letters and numbers should still be included.

Test that regex author classifier with super-dislike score wins

Test that regex author classifier matches patterns

Test that regex author classifier returns 0 when pattern doesn't match

apply_classifier_authors should ignore regex classifiers (they're scored separately)

Super downvote author classifier returns -2.

Super downvote tag classifier returns -2.

Super downvote tag (-2) wins even when another tag is positive (+1).

Text classifiers should match visible text, ignoring inline HTML tags.

Super downvote title classifier returns -2.

Test that regex URL classifier with negative score works

Test that regex URL classifier matches patterns

Test that regex URL classifier returns 0 when pattern doesn't match

Test that exact URL classifier is case-insensitive

Test that exact URL classifier matches as substring (case-insensitive)

Test that exact URL classifier returns 0 when no match

Test that exact URL classifiers require Premium tier

Each FetchArchiveFeedsChunk in the chord should have an increasing countdown delay.

The archive fetch should create a Celery chord with a FinishFetchArchiveFeeds callback.

The FinishFetchArchiveFeeds callback should have a time_limit matching MAX_SECONDS_COMPLETE_ARCHIVE_FETCH so post-fetch sync_redis doesn't get

archive_page should create new archive.

Longer content should replace shorter content.

Shorter content should not replace longer content.

Revisiting soft-deleted archive should undelete it.

archive_page should update existing archive on revisit.

If the hub returns a 202 response, we should not assume the subscription is verified.

Balanced quotes should pass through unchanged.

Tests that 1 + 1 always equals 2.

Tests that 1 + 1 always equals 2.

Tests that 1 + 1 always equals 2.

Tests that 1 + 1 always equals 2.

Tests that 1 + 1 always equals 2.

Basic word search should return matching stories.

Invalid JSON should return error.

Batch ingest should reject >100 items.

Missing URL in batch item should result in error for that item.

Batch ingest should create multiple archives.

Batch ingest should handle partial failures.

Batch ingest should require authentication.

Blocked URLs should be skipped in batch.

Batch ingest should update stats correctly.

Banking domains should be blocked.

Checkout and cart pages should be blocked.

Email domains should be blocked.

Localhost URLs should be blocked.

Login pages should be blocked by pattern.