Rolino fixed changes

Brand deletion now clears Blog plan items before related project cascades, so brands with active Blog plans can be removed without a database constraint failure.

Blog Settings now keeps consistent bottom space after its final setup panel on desktop and mobile screens.

Publishing-integration setup actions now stay right-aligned, and the provider list uses the shared accessible scroll area.

Blog update-draft confirmations now use the full dialog width before long article titles wrap.

Blog plan items that need attention now explain the exact blocker and offer clear recovery actions. Update-draft adoption is limited to the current active plan revision and its live update target, restored ideas clear stale decision metadata, and stale review choices refresh to the current state.

The landing page channel list now shows the X mark once while keeping its accessible name.

Blog combined approval now preserves an article's explicitly selected publication time and uses the editorial-plan time only when no exact time was saved.

Weekly Blog site refreshes now stop after their bounded retry policy instead of reopening the same failed job on every automation run.

Blog quality checks now match complete forbidden terms and phrases without rejecting unrelated words that contain the same letters.

Blog article review now keeps clear bottom space below the exact revision approval card.

Blog revision history now opens without a React form-action warning, and each restore action still selects the correct revision.

Blog article review now loads its featured image eagerly when it is the page's largest visible element, which improves initial rendering and removes the Next.js performance warning.

Blog successor plans now wait until the current plan ends and every item is dismissed or represented by an article, so unfinished planned work is not abandoned.

Blog publication now remains successful when observational checkpoint enrollment fails, and later retries can restore the missing measurement jobs.

Blog performance observations now match Analytics landing pages and Search Console pages through provider-aware paths, include bounded site-wide context, and show the recorded source metrics in the article review.

Agent Blog reads now include bounded measurement checkpoints and successor-plan progress without exposing provider-private data.

Blog automation now preserves an explicitly confirmed schedule instead of converting it into immediate publication, replays combined approvals safely, rotates bounded site scans fairly, preserves successful article revisions when image preparation fails, enrolls every measurement checkpoint durably, keeps historical observations stable, and exposes failed successor-plan recovery.

Backlink discovery now keeps useful results when product loading, query planning, or one search stage fails, and records the failed stage for diagnosis.

Backlink fallback searches now prefer verified editorial contexts over product-category and transformation phrases that mainly return competitors, and discard Lemon8 social posts before page fetching.

Strong complementary roundups can now pass without an existing external link, and exact backlink evidence survives harmless page whitespace and zero-width formatting differences. Evidence-mismatch pages can be checked again after this correction.

Backlink review now accepts verified submission pages and strong agency or ecosystem articles when the publisher promotes a different, non-competing service. Pages blocked only by an older server policy can be checked again, and generic Homepage target labels no longer become search topics.

Extended backlink searches now use target-page topics and new publisher angles when query planning falls back, instead of repeating the same fixed searches and returning the same narrow result pool.

Backlink discovery now accepts strong articles from trusted professional and educational publishers, plus safe curated ecosystem roundups, even when the page does not already contain an external link; competitors and commercial conflicts remain excluded.

Complementary ecosystem articles can now qualify when they have a strong natural placement and realistic outreach path, even when the fetched page does not already contain an external link.

The competitor-selection gate now runs even when adaptive query planning falls back, and every model-approved page records whether the server gate or exact-evidence check prevented it from being saved.

Backlink page fetching now uses a practical cache and the provider-recommended timeout window, and relevant agency or educational articles can pass the final eligibility gate when they have a natural placement.

Sol reasoning requests now omit incompatible sampling controls, allowing backlink query planning and verified-page analysis to use a compatible OpenRouter endpoint.

Backlink analysis now uses smaller, bounded page batches in parallel, keeps successful batches when one fails, and gives cost-conscious research models enough time to finish.

Backlink discovery now rejects articles whose audience or editorial point of view does not match the target product, even when the page shares broad topic words.

Backlink discovery now checks each publisher's homepage and rejects product landing pages and vendor-authored roundups for products similar to the backlink target.