P2: News block remote fetching — batch requests, add timeouts, fix cache keys, cap pagination

[Herm71]

Roadmap items 14, 15, 16, 19 (P2). Grouped: one caching/fetching pass over the remote News path. 14 and 16 are verified.

The News block is the only block that fetches over REST from the news site (News_Request, base URL chosen from wp_get_environment_type()). Its fetch path currently has four compounding problems.

Roadmap item 14 — N+1 blocking HTTP on cold render

One wp_remote_get per featured image plus one per coauthor, on top of the posts request itself — up to ~19 sequential blocking calls to render a 9-post block with a cold cache.

Fix: request _embed on the posts request so media and author data arrive in one response; drop the per-object follow-up calls.

📍 src/Components/News_Block_Controller.php:132-190

Roadmap item 15 — No timeout on any remote request

Every wp_remote_get uses the 5s default. Combined with 14, a slow or unreachable news host can stall a front-end page render for tens of seconds.

Fix: set an explicit short timeout (2–3s), and make a failed fetch degrade to an empty/cached result rather than blocking.

📍 src/Request/News_Request.php:17

Roadmap item 16 — Cache keys embed taxonomy_ids

Media and coauthor transients are keyed with the block's taxonomy_ids, so the same attachment or author is cached separately for every block configuration that references it. Cache hit rate collapses; transient table grows.

Fix: key per-object caches by object ID only. Keep the taxonomy-scoped key for the posts query result, where it belongs.

📍 src/Components/News_Block_Controller.php:125-130

Roadmap item 19 — Unbounded pagination recursion

News_Request::request() recurses over paginated results with no page cap. A large X-WP-TotalPages means unbounded sequential fetches on a single render.

Fix: cap the page count (and/or stop once enough posts for the block have been collected).

📍 src/Request/News_Request.php:30-40

Acceptance

• Cold render of a 9-post News block makes a small bounded number of HTTP requests (target: 1–2)
• Explicit timeout on every remote call; unreachable host degrades gracefully instead of hanging the render
• Per-object transients shared across block configurations
• Hard cap on pages fetched per request
• Measured before/after request count and wall time noted on the PR