Summary

This PR introduces two major improvements:

1. Block Fetch Prefetch Pipeline

Adds a configurable prefetch mechanism that decouples block fetching from validation processing, allowing the network pipeline to stay full while validation processes earlier batches. This significantly improves catchup throughput.

Key changes:

• New FetchPrefetchDepth setting (default: 5 batches) to control how far ahead fetching runs
• Implements batchFetchWithPrefetch() and distributeBatchesToWorkers() for pipeline architecture
• Uses buffered channel to naturally throttle fetches when validation falls behind
• Falls back to original synchronous behavior when prefetchDepth=0
• Adds detailed logging for pipeline monitoring

2. Improved Shutdown Consistency

Enhances graceful shutdown behavior across GRPC and Kafka consumer to prevent hangs and race conditions.

GRPC improvements (util/grpc.go):

• Adds 5-second timeout for graceful shutdown
• Falls back to forced Stop() if graceful shutdown hangs
• Better logging for shutdown states

Kafka consumer improvements (util/kafka/kafka_consumer.go):

• Fixes race condition where signal handler could attempt cancellation after context already done
• Properly checks context state before cancelling

Subtree validation improvements (services/subtreevalidation/check_block_subtrees.go):

• Adds context cancellation checking between merkle tree levels
• Allows faster shutdown during long-running block validation

Test Plan

• Run catchup performance tests to verify improved throughput with prefetch enabled
• Test graceful shutdown behavior under various load conditions
• Verify no regressions in block validation correctness
• Test with prefetchDepth=0 to ensure backward compatibility
• Verify GRPC shutdown doesn't hang with stuck connections
• Test Kafka consumer shutdown under signal interrupts

🤖 Generated with Claude Code