feat: Worker improvements and Run Now duplicate prevention

- Fix Run Now to prevent duplicate task creation
- Add loading state to Run Now button in UI
- Return early when no stores need refresh
- Worker dashboard improvements
- Browser pooling architecture updates
- K8s worker config updates (8 replicas, 3 concurrent tasks)

backend/docs/_archive/WORKER_TASK_ARCHITECTURE.md

@@ -504,6 +504,103 @@ The Workers Dashboard shows:
 | `src/routes/worker-registry.ts:148-195` | Heartbeat endpoint handling |
 | `cannaiq/src/pages/WorkersDashboard.tsx:233-305` | UI components for resources |
 
+## Browser Task Memory Limits (Updated 2025-12)
+
+Browser-based tasks (Puppeteer/Chrome) have strict memory constraints that limit concurrency.
+
+### Why Browser Tasks Are Different
+
+Each browser task launches a Chrome process. Unlike I/O-bound API calls, browsers consume significant RAM:
+
+| Component | RAM Usage |
+|-----------|-----------|
+| Node.js runtime | ~150 MB |
+| Chrome browser (base) | ~200-250 MB |
+| Dutchie menu page (loaded) | ~100-150 MB |
+| **Per browser total** | **~350-450 MB** |
+
+### Memory Math for Pod Limits
+
+```
+Pod memory limit:     2 GB (2000 MB)
+Node.js runtime:      -150 MB
+Safety buffer:        -100 MB
+────────────────────────────────
+Available for browsers: 1750 MB
+
+Per browser + page:    ~400 MB
+
+Max browsers: 1750 ÷ 400 = ~4 browsers
+
+Recommended: 3 browsers (leaves headroom for spikes)
+```
+
+### MAX_CONCURRENT_TASKS for Browser Tasks
+
+| Browsers per Pod | RAM Used | Risk Level |
+|------------------|----------|------------|
+| 1 | ~500 MB | Very safe |
+| 2 | ~900 MB | Safe |
+| **3** | **~1.3 GB** | **Recommended** |
+| 4 | ~1.7 GB | Tight (may OOM) |
+| 5+ | >2 GB | Will OOM crash |
+
+**CRITICAL**: `MAX_CONCURRENT_TASKS=3` is the maximum safe value for browser tasks with current pod limits.
+
+### Scaling Strategy
+
+Scale **horizontally** (more pods) rather than vertically (more concurrency per pod):
+
+```
+┌─────────────────────────────────────────────────────────────────────────┐
+│ Cluster: 8 pods × 3 browsers = 24 concurrent tasks                       │
+│                                                                          │
+│  ┌─────────────┐ ┌─────────────┐ ┌─────────────┐ ┌─────────────┐       │
+│  │ Pod 0       │ │ Pod 1       │ │ Pod 2       │ │ Pod 3       │       │
+│  │ 3 browsers  │ │ 3 browsers  │ │ 3 browsers  │ │ 3 browsers  │       │
+│  └─────────────┘ └─────────────┘ └─────────────┘ └─────────────┘       │
+│                                                                          │
+│  ┌─────────────┐ ┌─────────────┐ ┌─────────────┐ ┌─────────────┐       │
+│  │ Pod 4       │ │ Pod 5       │ │ Pod 6       │ │ Pod 7       │       │
+│  │ 3 browsers  │ │ 3 browsers  │ │ 3 browsers  │ │ 3 browsers  │       │
+│  └─────────────┘ └─────────────┘ └─────────────┘ └─────────────┘       │
+└─────────────────────────────────────────────────────────────────────────┘
+```
+
+### Browser Lifecycle Per Task
+
+Each task gets a fresh browser with fresh IP/identity:
+
+```
+1. Claim task from queue
+2. Get fresh proxy from pool
+3. Launch browser with proxy
+4. Run preflight (verify IP)
+5. Execute scrape
+6. Close browser
+7. Repeat
+```
+
+This ensures:
+- Fresh IP per task (proxy rotation)
+- Fresh fingerprint per task (UA rotation)
+- No cookie/session bleed between tasks
+- Predictable memory usage
+
+### Increasing Capacity
+
+To handle more concurrent tasks:
+
+1. **Add more pods** (up to 8 per CLAUDE.md limit)
+2. **Increase pod memory** (allows 4 browsers per pod):
+   ```yaml
+   resources:
+     limits:
+       memory: "2.5Gi"  # from 2Gi
+   ```
+
+**DO NOT** simply increase `MAX_CONCURRENT_TASKS` without also increasing pod memory limits.
+
 ## Monitoring
 
 ### Logs