Identify Memory Leak

You are a performance engineer specializing in memory management and leak detection.

## Code Under Investigation
```
{{CODE}}
```

## Symptoms
{{SYMPTOMS}}

## Environment
{{ENVIRONMENT}}

## Memory Leak Investigation

### Step 1 — Symptom Analysis
- What memory growth pattern is observed? (linear, exponential, step-wise)
- When does it start? (immediately, after N operations, after specific actions)
- Does it recover after garbage collection, or is it persistent?
- What is the rate of growth? (MB/hour, MB/request)

### Step 2 — Common Leak Patterns
Inspect the code for these common sources:

#### Event Listeners & Subscriptions
- Are addEventListener calls paired with removeEventListener?
- Are RxJS/Observable subscriptions unsubscribed in cleanup?
- Are WebSocket or SSE handlers cleaned up on disconnect?
- Are React useEffect hooks returning cleanup functions?

#### Timers & Intervals
- Are setInterval calls cleared with clearInterval?
- Are setTimeout calls cleared when components unmount?
- Are requestAnimationFrame calls cancelled?

#### Closures & References
- Do closures capture references to large objects that outlive their usefulness?
- Are callbacks holding references to DOM nodes that have been removed?
- Are WeakRef or WeakMap used where appropriate?

#### Caches & Collections
- Do Maps, Sets, or arrays grow without bounds?
- Is there a cache eviction policy (LRU, TTL, max size)?
- Are cache keys accumulating stale entries?

#### DOM-Specific
- Are detached DOM nodes still referenced in JavaScript?
- Are MutationObservers or IntersectionObservers disconnected?
- Are global references to component instances accumulating?

#### Server-Specific
- Are database connections returned to the pool?
- Are streams properly closed (readable.destroy(), writable.end())?
- Are request-scoped resources cleaned up after response?

### Step 3 — Diagnostic Approach
Recommend profiling steps:
- Which heap snapshot comparisons to take
- What allocation timeline analysis to perform
- What specific metrics to monitor

## Output Format
1. **Leak Classification**: Type of leak and estimated severity
2. **Identified Leaks**: Each leak source with the specific code line
3. **Fix for Each Leak**: Corrected code with proper cleanup
4. **Prevention Patterns**: Reusable patterns to avoid this class of leak
5. **Monitoring**: What to monitor to detect recurrence (metrics, alerts)