eth/protocols/eth: increase maxTxPacketSize from 100 KB to 1 MB

[lucca30]

Summary

Increases maxTxPacketSize in eth/protocols/eth/broadcast.go from the go-ethereum default of 100 KB to 1 MB to match Polygon mainnet's actual transaction throughput at high utilization.

Background: what maxTxPacketSize controls

This constant is the flush threshold used in two places:

1. Direct broadcast (broadcastTransactions): the loop accumulates full transactions and flushes when the accumulated size hits the limit. Only sqrt(peers) receive direct txs.
2. GetPooledTransactions responses: when a peer fetches full txs after receiving hash announcements (the path used by the majority of peers), the response is also bounded by this limit.

Announcements (NewPooledTransactionHashesPacket) are unaffected — they carry only 32-byte hashes, so even a 440-tx block produces ~14 KB, negligible against any reasonable limit.

Problem: 100 KB is severely undersized for Polygon

Sampling 11 Polygon mainnet blocks at high utilization (blocks 84665222–84665232, sampled 2026-03-25):

Block	Txs	Gas Used	Gas Util	Tx Data	@ 1 MB
84665222	277	86.7M	72.2%	365.0 KB	1 msg
84665223	249	60.9M	50.7%	261.4 KB	1 msg
84665224	374	116.6M	97.1%	505.4 KB	1 msg
84665225	437	120.0M	100.0%	520.6 KB	1 msg
84665226	351	118.3M	98.6%	510.2 KB	1 msg
84665227	318	110.2M	91.9%	458.9 KB	1 msg
84665228	443	99.3M	82.8%	426.9 KB	1 msg
84665229	278	86.2M	71.8%	351.2 KB	1 msg
84665230	305	93.2M	77.7%	391.4 KB	1 msg
84665231	317	113.1M	94.3%	480.0 KB	1 msg
84665232	244	111.5M	92.9%	460.1 KB	1 msg

Average: 430 KB/block. Max observed (100% utilization): 520 KB.

At 100 KB, a full block requires 5+ sequential p2p messages per peer. The impact compounds on the fetch path: the majority of peers receive only hash announcements and call GetPooledTransactions to fetch full txs. A 100 KB response limit means multiple round trips to converge the mempool across the full peer set — not just the small sqrt(peers) direct broadcast set.

Why not 512 KB?

512 KB was initially considered, but full blocks already reach ~520 KB — leaving essentially no margin. Blocks 84665224, 84665225, and 84665226 all exceed or come within 2 KB of 512 KB at 97–100% utilization. Any sustained period at full capacity would regularly split into 2 messages.

Size selection: 1 MB

Limit	Covers full block (~520 KB)	Headroom to 200M gas limit (~870 KB)	% of p2p max (16 MB)	vs witness (15 MB)
100 KB (current)	No — 5+ messages	No	0.6%	0.7%
512 KB	Barely — full blocks exceed it	No	3.2%	3.4%
1 MB	Yes — 2x headroom	Yes — fits in 1 message	6.3%	6.7%
2 MB	Yes — 4x headroom	Yes	12.5%	13.3%

1 MB is chosen because:

• It covers the current maximum full-block payload (~520 KB) with ~2x headroom
• It absorbs a gas limit increase to ~200M gas (producing ~870 KB blocks) without another bump
• It stays at 6.3% of the 16 MB p2p hard limit
• It is well below the ~15 MB witness messages already exchanged on this network

Network topology consideration

A significant portion of Polygon's transaction volume originates from a small number of high-throughput RPC providers (Alchemy, Infura, QuickNode, etc.) that submit large batches of transactions in near every block. These nodes tend to be well-connected to validators and their txs propagate via direct broadcast. However, the broader peer set — archive nodes, indexers, regional RPC nodes — relies entirely on the announce→fetch path. With a 100 KB response limit, these nodes consistently need 5+ round trips per block to converge their mempools, degrading local tx visibility and fee estimation accuracy.

Risk assessment

• Protocol compatibility: none. Packet size limits are local to the sender; receivers handle any size up to the p2p hard max.
• Memory: at 50 peers × 1 MB in-flight = 50 MB. Negligible on validator hardware.
• Head-of-line blocking: already accepted at 15 MB for witnesses. 1 MB is not a new risk category.
• Upstream divergence: go-ethereum targets Ethereum mainnet (~15 txs/12s block). This is an intentional Bor-specific tuning, documented in the constant comment.

Testing

Built and verified. The constant affects only the flush threshold in the broadcast/announce loops; no behavioral changes beyond batch sizing.

[claude]

Claude Code Review

This repository is configured for manual code reviews. Comment @claude review to trigger a review.

_{Tip: disable this comment in your organization's Code Review settings.}

[claude]

Code review

No issues found. Checked for bugs and CLAUDE.md compliance.

[claude]

Code review

No issues found. Checked for bugs and CLAUDE.md compliance.

Note: The PR title says "512 KB" but the code sets maxTxPacketSize = 1024 * 1024 (1 MB). The description correctly explains the 1 MB rationale — the title appears to be stale from an earlier iteration.

[sonarqubecloud]

Quality Gate passed

Issues
0 New issues
0 Accepted issues

Measures
0 Security Hotspots
0.0% Coverage on New Code
0.0% Duplication on New Code

See analysis details on SonarQube Cloud

[cffls]

LGTM, thanks!

[codecov]

Codecov Report

✅ All modified and coverable lines are covered by tests.
✅ Project coverage is 51.56%. Comparing base (0d6ee4b) to head (e9c1799).
⚠️ Report is 1 commits behind head on develop.

Additional details and impacted files

@@             Coverage Diff             @@
##           develop    #2160      +/-   ##
===========================================
- Coverage    51.57%   51.56%   -0.02%     
===========================================
  Files          882      882              
  Lines       154164   154164              
===========================================
- Hits         79505    79488      -17     
- Misses       69489    69509      +20     
+ Partials      5170     5167       -3     

Files with missing lines	Coverage Δ
eth/protocols/eth/broadcast.go	23.66% <ø> (ø)

... and 21 files with indirect coverage changes

Files with missing lines	Coverage Δ
eth/protocols/eth/broadcast.go	23.66% <ø> (ø)

... and 21 files with indirect coverage changes

🚀 New features to boost your workflow:

• ❄️ Test Analytics: Detect flaky tests, report on failures, and find test suite problems.
• 📦 JS Bundle Analysis: Save yourself from yourself by tracking and limiting bundle sizes in JS merges.