Cache JSON stats in memory instead of per-request file I/O

[lbedogni]

Problem
RequestHandler.handle_device_inference_result() reads device_inference_times.json and edge_inference_times.json on every request, then writes them back after EMA updates.

Impact

• Throughput bottleneck, especially with many concurrent devices.
• Increased latency per request.

Proposed Fix

• Keep the inference time dictionaries in memory (module-level or singleton).
• Flush to disk periodically (every N updates or X seconds) with atomic writes (write to temp file then rename).
• Consider file locking if multiple processes, or consolidate to single process.

Files: src/server/communication/request_handler.py Labels: performance, P2

[lbedogni]

⚠️ PARTIALLY RESOLVED

Current Status:

• ✅ 7 files currently use type hints (typing imports found)
• ❌ Majority of codebase still lacks type annotations

Files with type hints:

• src/client/python/delay_simulator.py
• src/server/core/delay_simulator.py
• src/server/core/variance_detector.py
• src/server/statistics/statistics_visualizer.py
• src/server/statistics/statistics_collector.py
• scripts/simulation/analyze_simulation.py
• scripts/analysis/variance_analysis.py

Remaining Work:

• Add type hints to remaining ~80+ Python files
• Focus on core modules: request_handler.py, http_client.py, edge_initialization.py
• Add mypy configuration for type checking
• Document type annotation style guide

Priority: Medium - type hints improve IDE support and catch bugs early

[lbedogni]

✅ Completed - see AGILE_ACTION_PLAN.md