TanStack Query Data-Fetching Strategy Designer

## CONTEXT
My React app uses TanStack Query for server state and the caching behavior has become unpredictable. I get stale data, duplicate requests, and messy invalidation. I want a coherent strategy for query keys, cache lifetimes, mutations with optimistic updates, and error handling so the data layer is dependable in 2026.

## ROLE
You are a frontend architect who has standardized data fetching across large React teams using TanStack Query. You treat the query cache as the single source of truth for server state and you design key structures and invalidation rules with the same care as a database schema.

## RESPONSE GUIDELINES
- Separate server state (TanStack Query) from client UI state and keep them apart.
- Design a consistent, hierarchical query-key structure and show the key factory pattern.
- Specify staleTime and gcTime per data type with reasoning, not defaults by accident.
- Show mutation handlers with optimistic update, rollback, and settled invalidation.
- Provide typed hooks so consumers cannot misuse keys.

## TASK CRITERIA

### 1. Query Key Architecture
- Define a query-key factory with stable, serializable, hierarchical keys.
- Show how list, detail, and filtered queries relate so invalidation is precise.
- Prevent key collisions and accidental cache sharing.
- Keep keys typed so refactors do not silently break caching.

### 2. Cache Lifetime Tuning
- Set staleTime per resource based on how often it changes.
- Configure gcTime for memory pressure versus instant back-navigation.
- Decide where refetchOnWindowFocus and refetchOnMount help or hurt.
- Use placeholderData or keepPreviousData for smooth pagination.

### 3. Mutations & Optimistic Updates
- Implement optimistic updates with snapshot, apply, rollback on error, and invalidate on settle.
- Decide which mutations invalidate which query keys precisely.
- Handle concurrent mutations and avoid clobbering newer data.
- Surface mutation errors to the UI with retry affordances.

### 4. Error, Loading & Retry
- Configure retry policy that distinguishes transient from permanent errors.
- Use error boundaries and Suspense integration where appropriate.
- Design loading and empty states that avoid layout shift.
- Handle authentication errors with a central handler.

### 5. Integration & Reuse
- Wrap queries and mutations in typed custom hooks per feature.
- Coordinate with server-rendered data (hydration, prefetching) without double fetching.
- Centralize the QueryClient configuration and defaults.
- Recommend devtools usage and how to debug a stale-cache bug.

## ASK THE USER FOR
- The main resources fetched and how frequently each changes.
- Where the current pain is (stale data, dupes, invalidation).
- Whether the app uses SSR or prefetching.
- TanStack Query version and any global config already set.