# Lean Proof Caps EIP-7002’s Block Call at 210,000 Gas

Published 2026-10-26 · Crypto.world Team · Crypto.world

> A Lean 4 proof shows that Ethereum’s EIP-7002 block-end system call succeeds under specified code and fork conditions and uses no more than 210,000 gas. A second result tracks committed withdrawal requests through the queue and into block outputs, preserving their order and multiplicity under the theorem’s assumptions.

210,000 gas is the ceiling a new Lean 4 proof assigns to Ethereum’s EIP-7002 system call, out of the 30 million gas available to it. The call runs at the end of every block; if it fails, the block is invalid. [Ethereum Research](https://ethresear.ch/t/lean-verification-of-eip-7002-system-call-success-and-fifo-requests/26180) poster skbaek published the result on October 26, alongside a second proof that tracks withdrawal requests from submission through the queue and into block outputs.

## The EIP-7002 call must not fail

EIP-7002 lets validator withdrawal credentials submit execution-layer requests for exits and partial withdrawals. Those requests pass through a 504-byte contract at `0x00000961Ef480Eb55e80D19ad83579A64c007002`, written in geas assembly. The protocol calls the contract at each block’s end to collect requests.

The post describes the queue result this way:

> So no committed submission is ever lost, duplicated, reordered, or emitted before it was submitted.
>
> skbaek, Ethereum Research poster

The first theorem says the call succeeds for any contract storage state, provided the canonical runtime code is installed and the block uses one of the covered forks: Prague, Osaka, BPO1 or BPO2. It shows that the call’s gas cost is computed exactly from the state and never exceeds 210,000. The proof is stated against Jaune, an executable model of Ethereum’s EVM and block processing written in Lean.

## Up to 16 requests leave in order

The queue theorem compares three ordered lists: committed submissions, requests emitted in block outputs, and entries still in contract storage. It proves that submissions equal emitted requests followed by the remaining queue. Because the records include the caller, public key and amount, the comparison checks the full request and preserves both order and duplicates.

The contract emits up to 16 queued entries per block. For an entry at position *q* when a block computes its requests, the proof says it will be emitted after exactly floor(*q*/16) blocks, provided the history continues to meet the theorem’s assumptions. Later submissions cannot push that entry back; the result does not promise that Ethereum will produce those future blocks.

The authors say the proofs use only Lean’s three standard axioms: `propext`, `Classical.choice` and `Quot.sound`. The work follows four pre-deployment audits and a loop-bounded Halmos check, but addresses a different question: whether the deployed bytecode’s behavior matches these claims under the modeled block-processing rules.

## The fee caveat is in the bytecode

The FIFO result does not prove that every request pays the fee specified by EIP-7002. The EIP calculates its fee using unbounded integers, while the contract uses 256-bit arithmetic; if an intermediate calculation wraps, the contract’s fee can fall below the EIP’s. The post reports a machine-checked example of underpayment in a synthetic chain, but says that example does not show such an underpayment can occur on mainnet.

The proofs also do not establish that Jaune matches every behavior of Ethereum clients, nor do they authenticate a mainnet state root or prove a particular transaction was included. The history theorem has explicit premises, including conditions on the protocol’s system address and a bound on balances and request counts. It does not determine whether the consensus layer accepts a request for a given validator or when an exit or withdrawal completes.

## Bytecode proof, not an end-to-end exit guarantee

The findings concern the execution-layer contract’s call and queue behavior. They do not prove transaction-level inclusion or end-to-end liveness for a user’s request, and they do not extend to later forks beyond the four named in the post. The contract also retains dequeued storage slots, a point the authors identify as outside these proofs.

For EIP-7002, the concrete result is twofold: the modeled block-end call succeeds under the stated code and fork conditions, and committed requests leave the contract in order, with no request dropped, duplicated or emitted early within a qualifying history.

## Sources

1. Original source [Ethereum Research](https://ethresear.ch/t/lean-verification-of-eip-7002-system-call-success-and-fifo-requests/26180) ethresear.ch

## Related stories

- [X100 Attackers Extract $4.85M in USDT and 550,000 X](https://crypto.world/news/x100-exploit-485m-usdt-550000-x)
- [zkLink Nova Bridge Exploit Drains About $786,000](https://crypto.world/news/zklink-nova-bridge-exploit-786k-forged-batch-root)
- [Researchers Find 4.67M Malicious EIP-7702 Delegations](https://crypto.world/news/researchers-find-4-67m-malicious-eip-7702-delegations)
- [Researchers Test Proof-of-Reserves for Finding Exchange Wallets](https://crypto.world/news/researchers-test-proof-of-reserves-exchange-wallets)
- [Ethereum Draft Would Permanently Bar EIP-7702 Delegation](https://crypto.world/news/ethereum-draft-permanent-eip-7702-delegation-opt-out)

## Explore

[More articles](https://crypto.world/news)

---

Canonical: https://crypto.world/news/lean-proof-eip-7002-withdrawal-queue
