SocialAccountObserver::notifyOnboarding() lazy-loads workspace — crashes when 2+ SocialAccount rows are batch-hydrated

[paulocastellano]

Problem

App\Observers\SocialAccountObserver::notifyOnboarding() reads $socialAccount->workspace?->account without workspace being eager-loaded on the triggering model.

This is latent almost everywhere because Laravel only flips preventsLazyLoading on a hydrated batch when Builder::hydrate() loads more than one distinct row in the same query (Illuminate\Database\Eloquent\Builder::hydrate(), gated by count($items) > 1). A single-account code path never trips it. Model::shouldBeStrict(! $this->app->isProduction()) is set in app/Providers/AppServiceProvider.php:217, so:

• Local/testing: strict mode is on → any caller that batch-processes 2+ distinct SocialAccount rows and triggers a status-changing update() (which fires the observer) throws LazyLoadingViolationException immediately.
• Production: strict mode is off → the same code path degrades to a silent N+1 query instead of a crash.

Where this was found

Discovered while building App\Jobs\VerifyUpcomingPostConnections (PR for the proactive upcoming-post connection check). That job intentionally groups at-risk post_platforms by social_account_id and can call SocialAccount::markAsTokenExpired() on multiple distinct accounts within a single run — exactly the "2+ distinct rows hydrated, one triggers the observer" shape that surfaces the bug.

Fixed at the call site for that job only (app/Jobs/VerifyUpcomingPostConnections.php, atRiskPostPlatforms()): eager-loads socialAccount.workspace before any account in the batch can be updated. That closes the gap for this one job, not for the observer itself.

Why this needs a real fix, not another per-caller workaround

Any other current or future code path that batch-updates 2+ distinct SocialAccount rows to a different status without separately remembering to eager-load workspace will hit the same crash (locally/in tests) or silent N+1 (in production). Audited during the review of the job above — none of the other current call sites are exposed today (each operates on a single account per invocation), but that's incidental, not structural:

• app/Jobs/RefreshSocialToken.php — single account per job, not exposed today
• app/Jobs/PublishToSocialPlatform.php — single account per job, not exposed today
• app/Http/Controllers/Auth/*Controller.php (OAuth connect flows) — single account per request, not exposed today
• app/Jobs/VerifyWorkspaceConnections.php — already eager-loads workspace.owner (:34), not exposed

Suggested fix

Fix SocialAccountObserver::notifyOnboarding() itself so it doesn't depend on every caller remembering to eager-load workspace — e.g. load it inside the observer if missing ($socialAccount->loadMissing('workspace')), or restructure so the observer doesn't need the relation lazily at all.

Repro (local/testing)

Any test/factory flow that creates 2+ distinct SocialAccount rows in one query result and then triggers a status update() on more than one of them within that same hydrated batch reproduces LazyLoadingViolationException at the notifyOnboarding() call site.