error-boundary-strategist

Error Boundary Strategist

Design layered React error handling with Error Boundaries, error reporting, recovery UI, retry patterns, and graceful degradation so apps never show a white screen.

Decision Points

Error Severity Matrix & Routing Logic

ERROR TYPE                  SEVERITY      ACTION
─────────────────────────  ────────────  ─────────────────────────
Component render error     Critical      Route-level boundary → full page fallback
Widget/feature crash      Recoverable   Feature-level boundary → inline retry
Network timeout           Recoverable   Query error callback → toast + retry
Chunk load failure        Critical      App-level boundary → full reload
Type error in handler     Recoverable   Try/catch → log + continue
Memory leak crash         Critical      Global error → browser reload

Decision Tree:

Is error from React render tree?
├── YES: Error boundary will catch it
│   └── Is error in isolated widget?
│       ├── YES: Use feature-level boundary → show inline fallback
│       └── NO: Use route-level boundary → show page fallback
└── NO: Manual handling required
    └── Is error in async operation?
        ├── YES: Use query.onError or try/catch → show toast
        └── NO: Use window.onerror → report to Sentry

Boundary Placement Strategy

If component tree depth > 3 levels:
  → Add feature-level boundaries around each major widget

If route has > 5 async data sources:
  → Add route-level boundary + individual query error handling

If app has > 10 routes:
  → Add app-level boundary as ultimate fallback

Always:
  → Add global-error.tsx in Next.js apps

Failure Modes

Infinite Reset Loop

Detection: componentDidCatch fires repeatedly on same component within 30 seconds Symptom: User clicks "Try Again" → immediate re-crash → fallback shows again Root Cause: onReset doesn't clear corrupted state that caused original error Fix: Clear query cache, local state, and props that triggered error in onReset handler

White Screen of Death

Detection: No error boundary catches error → blank page with no UI Symptom: App shows empty white screen, no fallback UI visible Root Cause: Error boundary placed too low in tree or missing app-level catch-all Fix: Add error boundary wrapping entire app as ultimate safety net

Error Boundary Blindness

Detection: Async errors logged to console but not caught by boundaries Symptom: Promise rejections, event handler crashes slip through boundaries Root Cause: Error boundaries only catch render-phase errors, not async Fix: Use try/catch in handlers + query.onError for async operations

Context Poisoning

Detection: Error boundary catches error but corrupted context still exists Symptom: Reset works initially but subsequent actions crash due to bad context Root Cause: Context provider above boundary has corrupted state Fix: Reset context state in boundary's onReset or move boundary above context

Fallback Cascade Failure

Detection: Error fallback UI itself throws error → nested error boundaries trigger Symptom: Fallback component crashes → shows another fallback → potential infinite loop Root Cause: Complex fallback components with their own dependencies/logic Fix: Keep fallback UI extremely simple → static JSX with minimal dependencies

Worked Examples

Example: E-commerce Dashboard with Widget Isolation

Scenario: Dashboard with revenue chart, recent orders, and user stats. Chart widget crashes due to malformed API data.

Expert Approach:

1. Assess impact: Chart crash should not kill orders widget or navigation
2. Choose boundary level: Feature-level around chart widget (not route-level)
3. Design fallback: Show "Chart temporarily unavailable" with retry button
4. Reset strategy: Clear chart query cache, don't reset entire dashboard state

// Expert places boundary precisely around volatile widget
<ErrorBoundary
  fallbackRender={({ error, resetErrorBoundary }) => (
    <div className="chart-error">
      <p>Revenue chart temporarily unavailable</p>
      <button onClick={resetErrorBoundary}>Reload Chart</button>
    </div>
  )}
  onReset={() => {
    // Only clear chart-related state
    queryClient.removeQueries({ queryKey: ['revenue-chart'] });
  }}
  isolate  // This error won't bubble up to route level
>
  <RevenueChart />
</ErrorBoundary>

// Orders widget remains functional
<RecentOrders />  // No error boundary needed if stable

Novice Mistake: Places one boundary around entire dashboard → chart error kills everything Expert Insight: Isolate volatile components; stable components don't need boundaries

Testing the Setup:

// In development, inject errors to verify boundaries work
{process.env.NODE_ENV === 'development' && (
  <ErrorTrigger componentName="RevenueChart" />
)}

Example: Next.js App Router Error Hierarchy

Setup Navigation:

• /dashboard route with error.tsx
• /dashboard/analytics with its own error.tsx
• Global error.tsx as final fallback

Error Routing Logic:

Analytics widget crashes
  ↓
dashboard/analytics/error.tsx catches it
  ↓ (if error.tsx itself crashes)
dashboard/error.tsx catches it
  ↓ (if that crashes too)
global-error.tsx replaces entire page

Implementation:

// app/dashboard/analytics/error.tsx
'use client';
export default function AnalyticsError({ error, reset }) {
  return (
    <div role="alert">
      <h2>Analytics Error</h2>
      <p>Unable to load analytics data</p>
      <button onClick={reset}>Retry</button>
      <Link href="/dashboard">Back to Dashboard</Link>
    </div>
  );
}

// Sentry reporting in each error boundary
useEffect(() => {
  Sentry.withScope((scope) => {
    scope.setTag('boundary.level', 'analytics');
    scope.setContext('route', { path: '/dashboard/analytics' });
    Sentry.captureException(error);
  });
}, [error]);

Quality Gates

• App-level error boundary catches all unhandled React errors
• Route-level boundaries isolate pages from each other
• Feature-level boundaries isolate widgets within pages
• Every error fallback has working retry/reset functionality
• onReset handlers clear corrupted state that caused original error
• Async errors handled with try/catch + query error callbacks (not boundaries)
• Error reporting (Sentry/etc) captures all boundary errors with component stack
• Fallback UI uses semantic HTML (role="alert") for accessibility
• No error shows raw stack traces or technical messages to end users
• window.onerror and unhandledrejection handlers catch non-React errors
• Next.js apps have global-error.tsx as ultimate safety net
• Error boundaries tested by injecting crashes in development mode

NOT-FOR Boundaries

This skill should NOT be used for:

• Form validation errors → Use form-validation-architect for field-level validation and user-correctable errors
• HTTP status codes (400, 404, 500) → Use api-architect for structured API error handling
• Authentication failures → Use auth patterns for login/permission errors
• Business logic validation → Use domain-specific validation patterns
• Loading states → Use Suspense boundaries for async loading UI
• Network offline detection → Use connectivity-aware patterns

Delegation Rules:

• For field validation: "Email is invalid" → form-validation-architect
• For API errors: "Resource not found" → api-architect
• For auth: "Session expired" → auth error handling
• For loading: "Fetching data..." → Suspense + loading UI patterns