# There Will Be Spam: Characterizing State-Invariant Transactions and Speculative MEV

Crypto.world

> Blockchains permanently store every transaction, and Layer-2 rollups such as Optimism and Base produce data far faster than Ethereum, inflating the storage that full nodes must maintain.

**arXiv**: 2607.24172

**Published**: 2026-07-27

**Authors**: Vabuk Pahari, Johnnatan Messias, Christof Ferreira Torres

## AI summary

## Overview

**Research area:** Blockchain systems security and measurement — specifically Layer-1/Layer-2 resource consumption, Maximal Extractable Value (MEV), and on-chain abuse (cs.CR).

**Technical level:** Intermediate. Readers need basic familiarity with transactions, gas, rollups, and MEV, but the paper's core idea (transactions that leave the ledger unchanged) is explained from first principles.

**Scope:** A large-scale measurement of "state-invariant" transactions — on-chain transactions that change nothing except fee payments — across Ethereum, Optimism, and Base from September 1, 2023 to July 31, 2025, including classification of their causes.

## What This Paper Is About

Blockchains permanently store every transaction, and Layer-2 rollups such as Optimism and Base produce data far faster than Ethereum, inflating the storage that full nodes must maintain. The authors argue that a large share of this activity is "spam": transactions that execute code but leave the blockchain state unchanged apart from fees, so removing them would yield an identical ledger. The paper formally defines these as *state-invariant transactions*, measures how many exist across three chains, and identifies what produces them — chiefly speculative MEV, but also phishing and inscriptions.

## Key Contributions

1. **A formal definition of state-invariant transactions** — a systematic, chain-agnostic metric for measuring spam on EVM-compatible blockchains, defined as transactions whose removal leaves the resulting blockchain state identical, ignoring the transfer of fees to the block proposer.
2. **The first large-scale measurement** of such transactions on Ethereum, Optimism, and Base over September 1, 2023 to July 31, 2025, covering 5,004,526 Ethereum blocks and 30,240,000 blocks each on Optimism and Base, and identifying nearly 1.4 billion state-invariant transactions.
3. **A scalable, trace-independent method for classifying MEV bots** as speculative, non-speculative, or hybrid, based on whether precomputed token amounts appear in transaction calldata — extending prior work by showing speculative MEV includes liquidations, not just arbitrage.
4. **A breakdown of non-MEV spam sources**, showing that a substantial fraction of state-invariant transactions come from reverting transactions (e.g., slippage protection), address poisoning, and on-chain message passing such as inscriptions.

## Main Findings

- **State-invariant transactions are rare on Ethereum but dominant on L2s:** 22,830,229 on Ethereum (approximately 2.6% of all Ethereum transactions in the dataset), 122,454,486 on Optimism (approximately 24.4%), and 1,228,926,146 on Base (37.5%). Base alone accounts for over 1.2 billion.
- **Fees paid by state-invariant transactions:** 5,141.6 ETH on Ethereum, 894.7 ETH on Optimism, and 7,110.45 ETH on Base across the study period.
- **Failed transactions dominate on Ethereum but not on L2s:** approximately 69% of Ethereum state-invariant transactions are failed (tx.status = 0), versus 19% on Optimism and 32% on Base. Reverted transactions alone are 57.4% on Ethereum, 17.5% on Optimism, and 30.9% on Base; out-of-gas failures are 11.7%, 1.9%, and 1.7% respectively.
- **Failed transactions pay disproportionately:** despite being a minority on L2s, failed transactions account for 88.3%, 80.9%, and 67.7% of priority fees paid by state-invariant transactions on Ethereum, Optimism, and Base.
- **Successful state-invariant transactions are the majority on L2s:** 80.6% on Optimism and 67.4% on Base, versus 30.9% on Ethereum. Of these, 73.8% on Optimism and 62.7% on Base emit no logs at all.
- **Speculative MEV is the dominant source on L2s but not the only one:** it accounts for 57% of state-invariant transactions on Optimism and 68% on Base — contradicting the prior assumption that spam is exclusively arbitrage-driven.
- **Speculative MEV is not the most profitable strategy once costs are counted:** when the fees incurred by state-invariant transactions are included, speculative MEV becomes unprofitable on average, and most speculative arbitrage bots on Optimism and Base turn out to be unprofitable.
- **Address poisoning is the biggest non-MEV source on Ethereum:** address-poisoning campaigns account for 53% of all non-reverted state-invariant transactions on Ethereum. Zero-value transfers alone are 973,477 transactions (4.3%) on Ethereum, 1,177,219 (1.0%) on Optimism, and 1,903,424 (0.2%) on Base.
- **A sharp post-Dencun shift on L2s:** after the March 2024 Dencun upgrade, state-invariant transactions consistently account for roughly 40–50% of daily transactions and daily gas consumed on both Optimism and Base, while Ethereum stays far lower throughout.
- **Storage burden is already visible:** a Reth archive node occupies approximately 3.5 TB for Ethereum, 3.9 TB for Optimism, and 9.0 TB for Base — even though Ethereum has run since 2015 while Optimism and Base launched in 2021 and 2023.

*Note: the provided text is truncated partway through Section 4.2.2, so per-category breakdowns beyond the figures above (for example, exact inscription shares and per-chain speculative-MEV transaction counts) are not available in the content reviewed.*

## Methodology in Plain English

The naive way to test the definition would be to delete a transaction, replay the block, and compare the resulting state — correct but far too slow at billions of transactions. Instead, the authors run archive nodes for all three chains using Reth and use the `trace_replayBlockTransactions` function, which returns a per-transaction "stateDiff" showing which accounts changed and how. A transaction counts as state-invariant if the only accounts touched are the sender and the fee recipient, and if the balance movements exactly match the fee.

The fee accounting differs by chain: on Ethereum the sender's balance must drop by the total fee while the block proposer gains the total minus the burned base fee. On Optimism and Base, fees split into three parts — base fee, sequencer (priority) fee, and L1 fee — each going to its own fee vault, so the checks are made against those vaults.

To separate speculative from non-speculative MEV, the authors first detect successful arbitrage and liquidation transactions using an extended version of Torres et al.'s detection techniques, which yields the pool addresses, token addresses, and token amounts involved. The key observation is that token amounts are dynamic and depend on live pool state, so a searcher cannot hard-code them — if a token amount from a successful MEV trade appears inside the transaction calldata, the searcher must have computed it off-chain and is therefore non-speculative. Contracts commonly hard-code pool and token addresses for gas savings, so addresses are not used for this judgment.

To avoid accidental substring matches, they only test token amounts whose hexadecimal length exceeds a threshold derived from the calldata length and a tolerated false-positive rate of 10⁻⁶ (one in a million). Matching then runs in milliseconds without expensive tracing. Bots are labeled speculative if all their observed MEV transactions are speculative, non-speculative if none are, and hybrid otherwise — a category that captures bots that precompute a minimum input amount off-chain and refine it on-chain, or that swap bytecode over time.

Address poisoning is identified through zero-value transfers (a `Transfer` event with no balance change) and transfers of fake or tampered ERC-20 contracts that emit events without touching state, using labels from Tsuchiya et al.'s public dataset; dust attacks are deliberately excluded because they produce real state changes. Inscriptions are found by looking for self-transactions (sender equals recipient) whose calldata begins with the prefix `0x646174613a`, the ASCII string `data:`.

## Why This Matters

**Impact on research:** The paper replaces a bot-centric, arbitrage-only notion of spam with a state-based definition that can be computed at chain scale and applied uniformly to any EVM chain. It also challenges a widely repeated assumption — that speculative MEV is the whole story and that it pays — by showing both other major sources and a negative average return once spam costs are included.

**Real-world applications:**

- **Rollup design:** Sequencers and protocol designers get a concrete lever — pricing, throttling, or rejecting state-invariant transactions — with measured evidence of how much blockspace and storage they consume.
- **Wallet safety:** Since 53% of non-reverted state-invariant Ethereum transactions are address poisoning, wallets could use this analysis to flag look-alike-address activity and encourage full-address verification.
- **Node operation and storage planning:** The archive-node sizes reported (3.5 TB / 3.9 TB / 9.0 TB) give operators concrete figures, and the paper links spam growth to the rising cost of running full nodes that underpin decentralization.
- **MEV infrastructure:** Evidence that L2s lack the bundling and revert-protection services that suppress speculative execution on Ethereum points to a market gap for private order flow and simulation-based inclusion on rollups.

**Industry relevance:** L2 operators, MEV searchers and builders, wallet providers, block explorers and analytics firms, and anyone bearing the cost of chain growth all have a stake in the finding that roughly a quarter to a third of L2 transactions may be removable without changing the ledger.

## Future Directions

- **Bring revert protection to L2s.** The paper attributes Ethereum's low spam rate partly to bundling services that simulate transactions before inclusion; whether an equivalent private-order-flow design can work with centralized sequencers and 2-second blocks is open.
- **Protocol-level mitigation policies.** What fee schedule, gas accounting rule, or inclusion policy would correctly penalize state-invariant transactions without harming legitimate use (for example, legitimate slippage-protection reverts)?
- **Extend and complete the measurement.** Address-poisoning labels cover only Ethereum and have a gap between blocks 20,208,026 and 21,565,106; extending labeled data to Optimism and Base would sharpen the non-MEV picture, and the study could be widened to further EVM chains and later time periods.
- **Robustness of the calldata heuristic.** The speculative/non-speculative classifier could mislabel bots that obfuscate or encrypt calldata, since token amounts would no longer be extractable — the authors note this is a limitation, leaving room for trace-based or hybrid verification.

## Target Audience

Blockchain systems researchers and graduate students studying MEV, rollups, or on-chain abuse; L2 protocol and sequencer engineers weighing transaction-pricing or spam-mitigation policy; MEV searchers and builders who want to know whether their strategies are actually profitable net of spam costs; and security analysts and wallet developers working on address-poisoning defenses. Readers without blockchain background will find the definition intuitive but will need the background section to follow the fee-vault and MEV details.

## Authors’ abstract

Blockchains rely on transparency and immutability to ensure trust, but these guarantees come at the cost of an ever-growing ledger that increasingly threatens decentralization by making it more expensive to store and maintain the full transaction history. In this work, we introduce state-invariant transactions, defined as transactions whose inclusion or removal does not affect the resulting blockchain state beyond transaction fees. We argue that these transactions constitute…

[The full abstract on arXiv](https://arxiv.org/abs/2607.24172)

[Read the original paper](https://arxiv.org/abs/2607.24172)

---

Canonical: https://crypto.world/paper/there-will-be-spam-characterizing-state-invariant-transactions-and-2607.24172
