Insufficient visibility into adaptive batching decisions impacts troubleshooting

BentoML's adaptive batching layer lacks visibility into how batching decisions are made, making it difficult to diagnose performance issues or understand batch behavior. Without debug logs showing batch size and latency per run, operators cannot correlate batching behavior with performance degradation.

Enable debug mode in RunnerApp to expose batch size and latency for each run. Monitor bentoml.runner.adaptive_batch.size histogram metrics (added in PR #2902) to track batching patterns. Check bentoml.runner.processing_latency and bentoml.runner.request.duration to correlate batch sizes with processing times.