Highlights
Fusaka devnet 3 continues to struggle with the non-finality test.
Tim Beiko has started a PR to update the Protocol Upgrade Process document that may speed up the process leading up to mainnet hard fork.
The Holešky testnet will be going offline soon after the Fusaka mainnet upgrade.
libp2p is planned to switch to community maintenance come September 30, 2025.
Multiple Reth nodes have stalled on block 23272427.
Public Calls
ACDT #51 – September 1, 2025
Fusaka devnet 3 status update – still struggling with non-finality test
Plan is to keep monitoring devnet 3 until finality is regained and launch devnet 5 (next week?) when fixes are merged in, and do the split chain again on devnet 3 with prior notice
Gas limit updates
Over the last week, testing stateful scenarios over any existing database and any existing chain
60 million will require fixes from Besu, decisions and fixes from geth about modexp improvement pre-Fusaka, and some stateful testing
Marius from geth mentions they have a PR that improves the gas computation and they want to upstream code to the Go standard library; will update next week
Glamsterdam updates
BALs – Felipe mentions that the spec side is passing CI and the testing side is talking to it
ePBS – Terence reports on work done on spec tests
New release for consensus specs – v1.6.0-alpha.6
ACDC #164 – September 4, 2025
Fusaka devnet 3
Barnabas from EthPandaOps published a doc sharing syncing bugs
Plan - merge all fixes to trunk (ideally by Monday) across clients, get devnet 3 to ~100% participation, run another non-finality test, and then look towards launching devnet 5 with 1500 nodes.
Following up on a discussion from ACDE last week about the Protocol Upgrade doc
Is it possible to move faster than the doc specifies? How do we strike a balance between speed and safety?
Tim Beiko made a PR to the doc in consultation with L2, LSTs, infra providers
At least 14 days between client releases and first testnet
At least 10 days between each test net
At least 30 days between client releases and mainnet
Lighthouse team generally in favor
Justin Florentine from Besu noting a lack of discourse regarding the PR from L2s, app developers, staking providers specifically regarding testnets' use as staging environments before going to prod
Reminder for Glamsterdam – next step for EIPs is at date of Fusaka mainnet releases
Marcin Sobczak from Nethermind brought attention to an issue appearing while testing EIP-7883; discussion on Telegram/Eth R&D Discord
Ethereum News & Discussions
The Ethereum Foundation made a new post regarding the sunsetting of the Holešky testnet. To quote the post, the “vast majority of remaining validator nodes will be shut down 2 weeks after the Fusaka upgrade has finalized on Holešky”.
Thomas Coratger shared the leanSpec GitHub repository, aimed at the "simplest possible Lean Ethereum Consensus specification".
leanConsensus call #7 took place on September 5. Will Corcoran posted a summary, highlighting
leanEthereum's components
PQ Interop's objectives
leanSpec template
leanMultiSig spec deep dive
leanVM, a minimal zkVM for leanEthereum
Editor's note: Check out https://leanroadmap.org/ to learn more about leanConsensus!
Due to difficulties resourcing, the Shipyard team plans to continue their maintenance of go-libp2p and js-libp2p only until September 30, 2025, after which these projects will transition to community maintenance. Notably, libp2p is a part of the networking stack of many blockchains, including Ethereum.
A number of Reth notes have stalled on a bad block. Georgios from Paradigm shared a post describing the incident and sharing recovery instructions for node operators.
EF Protocol's AMA on Reddit has concluded.
Joshua Rudolf shared notes on L2 Interop Working Group Call #13.
Jihoon Song and Milos Stankovic have worked on an implementation of EIP-7782 (6s slot times) via a post-Fusaka fork of geth/Prysm, and their prototype can now “remain stable over several epochs”.
nixo has published the September 2025 edition of State of EthMag.
Arvolear shared the reference implementation for ERC-8002.
EIP-2780 has been updated and proposed for Glamsterdam, aiming to increase maximum transactions per block.
Research Update
The following include LLM-generated summaries that may be slightly edited.
Payload Chunking by Toni Wahrstätter
Splitting Ethereum execution layer blocks into independently validated, fixed-gas “chunks” enables parallelized, stateless block validation and streaming, reducing latency and supporting future extensibility for sharding and zk-proofs.
* Post-Pectra Effects on Ethereum: Reorg Rate, Propagation, and Block Size by Seongwan Park
The Pectra hardfork and related Ethereum upgrades reduced reorg rates and block propagation delays by optimizing block and blob handling, decreasing block size variance, and shifting transaction data from calldata to blobs, thereby enhancing consensus stability and scalability.
An analysis of attestation timings in a 6-s slot by Maria Inês Silva
Reducing Ethereum slot times to 6 seconds is feasible, as block and attestation propagation consistently meet proposed deadlines, though validator misconfigurations and block content complexity remain significant factors influencing late attestations.
Gossipsub’s Partial Messages Extension and Cell Level dissemination by Marco Munizaga
Gossipsub’s Partial Message Extension enables efficient, cell-level dissemination of erasure-coded blobs in Ethereum by allowing nodes to request and propagate only missing data segments, significantly reducing bandwidth and latency without requiring consensus changes or a hard fork.
New EIPs/ERCs
No new EIPs or ERCs this week, though look above for updates concerning EIP-2780 and ERC-8002.
Client and Client-Related Updates
Besu released 25.8.0.
This is a required update.
Nimbus released v25.9.0.
go-ethereum released v1.16.3 “Impact Restrictors”.
Granding released 1.1.5.
mev-boost released v1.10-alpha3.
Lodestar released v1.34.0-rc.0.
Tooling, Languages & Libraries
Alloy released v1.0.30.
Solarity released v3.2.1.
Notably includes new
Bridge v0.1contracts for “simple cross-chain token transfers”.
Hardhat released v3.0.4, a bugfix release.
Foundry released v1.3.4, a bugfix release.
Rollup & EVM Alt-L1 News
Somnia Network, an EVM-compatible L1, is now live on mainnet.
Matt Huang, Founder of Paradigm, posted on X about Tempo, a new EVM-compatible blockchain incubated by Stripe and Paradigm
He followed up with a post justifying its design as an L1.
Starknet has released Grinta, aka 0.14.0. Whereas Starknet used to have a single sequencer, it now has three sequencers in consensus. Moreover, the new release includes preconfirmations, a mempool, and a fee market on
l2_gas.On September 4, Sonic posted that nodes must upgrade to 2.1 to avoid disconnection.