Stream legacy blocks straight to legacy without deserialization

[oskarszoon]

No description provided.

[github-actions]

🤖 Claude Code Review

Status: Complete

---

Current Review:

This PR optimizes legacy block streaming by avoiding deserialization overhead, which is good for CPU performance. However, there is one significant issue to address:

• Memory concern: RawBlockMessage uses io.ReadAll() to load entire blocks into memory (see comment). For 4GB blocks with multiple concurrent peers, this could cause memory pressure.

Positive changes:

• SubtreeProcessor atomic.Pointer migration is correctly implemented for thread safety
• Removed ConcurrentBlob disk caching eliminates I/O bottleneck for large blocks
• HTTP streaming timeout increased to 10min addresses the timeout issues
• Test coverage is comprehensive

The core optimization is sound, but consider the memory/CPU tradeoff for the io.ReadAll() approach vs true streaming.

[github-actions]

Memory Issue: Loading entire block into RAM

The io.ReadAll() call loads the entire block into memory. For 4GB blocks mentioned in the comments, this causes a 4GB+ memory spike per block request.

Problem:

• Multiple concurrent peer requests could cause severe memory pressure
• This defeats the streaming optimization mentioned in the PR description

Suggestion:
Consider whether the wire protocol supports true streaming without loading the entire payload into memory first, or if this memory tradeoff is acceptable for the CPU performance gain.

[oskarszoon]

The Bitcoin wire protocol message format requires:
[Magic: 4 bytes][Command: 12 bytes][Length: 4 bytes][Checksum: 4 bytes][Payload]

The checksum must be in the header BEFORE the payload. This means you need the entire payload to:

1. Calculate its length
2. Calculate its SHA256 checksum

We'll revisit this possible improvement later

[sonarqubecloud]

Quality Gate passed

Issues
1 New issue
0 Accepted issues

Measures
0 Security Hotspots
88.7% Coverage on New Code
1.5% Duplication on New Code

See analysis details on SonarQube Cloud