Fix: Meta Description: SEO plugin detection cache never expires and skips the "no plugin" case

[hbhalodia]

What?

Closes #970

Why?

1. The positive result is stored with no expiry and is only cleared on deactivated_plugin - a hook that's registered only while the Meta Description experiment is enabled. So the cache can go stale and never self-heal.
2. The "no SEO plugin active" result is never cached, so a full plugin scan runs on every call, including the front-end output_meta_description() path.

How?

• Updates the transient to cache it for a day.
• Update the activated, deactivated hook outside of experiment, so a cache would invalidate regardless of experiment is enabled or not.

Use of AI Tools

AI assistance: Yes
Tool(s): Claude Code
Model(s): Opus 5
Used for: Code, Unit Test under the direction provided by me. Implementation reviwed by me.

Testing Instructions

1. Activate a supported SEO plugin (e.g. Yoast) + enable the Meta Description experiment; trigger meta handling once to cache the slug.
2. Disable the Meta Description experiment.
3. Deactivate the SEO plugin.
4. Re-enable the experiment, This should use the plugin's meta key, i.e wpai_meta_description.

Changelog Entry

Fixed - Cached the active SEO plugin detection - with a TTL and immediate invalidation on any plugin activation/deactivation - so meta description meta-key lookups no longer re-scan active plugins on every request.

[github-actions]

The following accounts have interacted with this PR and/or linked issues. I will continue to update these lists as activity occurs. You can also manually ask me to refresh this list by adding the props-bot label.

If you're merging code through a pull request on GitHub, copy and paste the following into the bottom of the merge commit message.

Co-authored-by: hbhalodia <hbhalodia@git.wordpress.org>
Co-authored-by: dkotter <dkotter@git.wordpress.org>

To understand the WordPress project's expectations around crediting contributors, please review the Contributor Attribution page in the Core Handbook.

[codecov]

Codecov Report

❌ Patch coverage is 60.60606% with 13 lines in your changes missing coverage. Please review.
✅ Project coverage is 74.54%. Comparing base (8016220) to head (b482d0b).

Files with missing lines	Patch %	Lines
...des/Abilities/Meta_Description/SEO_Integration.php	55.55%	12 Missing ⚠️
includes/Admin/Upgrades/V1_4_0.php	80.00%	1 Missing ⚠️

Additional details and impacted files

@@              Coverage Diff              @@
##             develop     #973      +/-   ##
=============================================
- Coverage      74.57%   74.54%   -0.04%     
- Complexity      3132     3141       +9     
=============================================
  Files            132      133       +1     
  Lines          12213    12237      +24     
=============================================
+ Hits            9108     9122      +14     
- Misses          3105     3115      +10     

Flag	Coverage Δ
unit	74.54% <60.60%> (-0.04%)	⬇️

Flags with carried forward coverage won't be shown. Click here to find out more.

☔ View full report in Codecov by Harness.
📢 Have feedback on the report? Share it here.

🚀 New features to boost your workflow:

• ❄️ Test Analytics: Detect flaky tests, report on failures, and find test suite problems.
• 📦 JS Bundle Analysis: Save yourself from yourself by tracking and limiting bundle sizes in JS merges.

[dkotter]

Things look good to me, thanks! Holding on merging this for now though as we add a new upgrade routine here that has the file name hardcoded to v1.4.0. If we end up doing a minor release (v1.3.1 as an example) prior to releasing 1.4.0, this code will be inaccurate. Will merge this once we know for sure the version we'll be releasing this as

[dugyen]

Reviewed this in detail (diff, tests, and interaction with the surrounding code). This looks like a solid, correct fix — didn't find a blocking issue.

What it does right, matching both parts of #970:

• Adds a DAY_IN_SECONDS TTL to the positive cache in SEO_Integration.php, and now also caches a 'none' sentinel for "no SEO plugin active" so the front-end output_meta_description() path stops re-scanning every plugin on every request.
• Moves cache invalidation off Meta_Description::register() and onto an unconditional SEO_Integration::register_cache_invalidation() called from Main::load() (hooked to plugins_loaded), so the cache now self-heals via activated_plugin/deactivated_plugin regardless of whether the experiment is enabled — exactly the "invalidation independent of experiment toggle" fix the issue asked for.
• Goes a bit further than requested: the new V1_4_0 upgrade routine deletes the old non-expiring transient once on update, so sites with an already-stale infinite cache get a clean slate rather than waiting on the new logic. Also handles multisite network-wide activation by clearing the cache on every site in the network.
• Test coverage is thorough: positive/negative caching, TTL-independent invalidation on both hooks, the upgrade routine's version-gating (skips on fresh installs and already-upgraded sites), and multisite network-wide vs. single-site invalidation scope.
• I checked for stale references to the removed clear_active_plugin_cache() method/hook elsewhere in the codebase, including the pre-existing Meta_DescriptionTest.php — none found, so no regression there.

Two minor, non-blocking observations:

1. The multisite network-wide path (clear_cache_for_network()) fetches every site (get_sites(['number' => 0])) and does a switch_to_blog() round-trip per site on every plugin activate/deactivate, not just SEO-relevant ones. Fine for typical networks; could be slow on a network with many thousands of sites.
2. The phpcs:ignore comment on switch_to_blog() names a specific VIPMinimum sniff — worth confirming CI is happy with that exact sniff name for this repo's ruleset, but harmless either way.

Nice work, @hbhalodia.

---

AI assistance: Yes. Tool(s): Claude Code (Sonnet 5). Used for: reading the diff and tracing it against the rest of the codebase to write this review; I read and take responsibility for the assessment above.