feat(paper3): E1 censorship-resistance failover experiment (N=50, Sepolia testnet)#26
feat(paper3): E1 censorship-resistance failover experiment (N=50, Sepolia testnet)#26jhfnetboy wants to merge 1 commit into
Conversation
… Sepolia) Paper3 §5.1 E1 reproducible experiment: demonstrates that a hostile bundler (Relayer A, invalid API key → 401/403) cannot prevent a valid UserOperation from executing when an alternative permissionless bundler (Relayer B, Candide public) is available. AOA property verified: no new paymaster-server signature required after failover — eligibility is determined by on-chain state only. Results (Ethereum Sepolia, ERC-4337 v0.7, N=50): - 49/50 success (98%); 1 failure due to transient base-fee spike (Trial 30) - Mean Relayer A rejection latency: 1,215 ms - New paymaster signatures required: 0 - Representative on-chain anchor: block 10,895,210 Files: - packages/analytics/scripts/paper3_failover_simulation.mjs (experiment script) - packages/analytics/data/paper3_failover_results_2026-05-22T07-42-13.csv - packages/analytics/data/paper3_failover_summary_2026-05-22T07-42-13.json - .env.example (template — no real keys)
![chatgpt-codex-connector[bot]](https://avatars.githubusercontent.com/in/1144995?s=60&v=4){kind=small}
There was a problem hiding this comment.
💡 Codex Review
Here are some automated review suggestions for this pull request.
Reviewed commit: 7969ad97f2
ℹ️ About Codex in GitHub
Your team has set up Codex to review pull requests in this repo. Reviews are triggered when you
- Open a pull request for review
- Mark a draft as ready
- Comment "@codex review".
If Codex has suggestions, it will comment; otherwise it will react with 👍.
Codex can also answer questions or update the PR. Try commenting "@codex address that feedback".
| const rows = records.map(r => | ||
| Object.values(r).map(v => JSON.stringify(String(v))).join(',') |
There was a problem hiding this comment.
The CSV writer uses JSON.stringify for cell encoding, but JSON escaping (\") is not valid CSV escaping for embedded quotes. As soon as a field contains quotes or commas (for example relayerA_error with a JSON error payload), parsers split the row into extra columns and shift subsequent fields, corrupting experiment outputs and any downstream metrics derived from the CSV.
Useful? React with 👍 / 👎.
| if (resB.ok) { | ||
| userOpHash = resB.result; // eth_sendUserOperation returns UserOperationHash | ||
| success = true; | ||
| successCount++; | ||
| totalFailoverMs += latA; |
There was a problem hiding this comment.
A trial is marked successful immediately after eth_sendUserOperation returns, before confirming inclusion via eth_getUserOperationReceipt. This inflates successCount and meanFailoverLatencyMs when a bundler accepts into mempool but the operation is dropped, replaced, or never mined within the polling window, which can materially misstate the censorship-resistance result.
Useful? React with 👍 / 👎.
|
Closing: the E1 bundler-failover experiment has been removed from Paper3 §5.1. The failover property is generic ERC-4337 (not AOA-specific); P1 evidence is now based on code structural analysis of validatePaymasterUserOp. The failover simulation script and data files in this branch are no longer referenced in the paper. |
1 participant
Summary
Adds the reproducible E1 failover experiment for Paper3 §5.1 (SuperPaymaster AOA decentralization validation).
packages/analytics/scripts/paper3_failover_simulation.mjspackages/analytics/data/paper3_failover_results_2026-05-22T07-42-13.csv+_summary_*.json.env.example(no real keys committed)Experiment Design
0x0000000071727De22E5E9d8BAf0edAc6f37da032Results
txHash 0x1afff9d0...a0aaAOA Property Validated
Reproduction
Test plan
.env.examplecontains no real credentials