Glamsterdam Upgrade on Hoodi Testnet Postponed Indefinitely

Ethereum core developers postponed the Glamsterdam upgrade on the Hoodi testnet (epoch 132,352, previously 2026-10-26 17:42:48 UTC) with no replacement date, after EIP-8037/8038 gas repricing was found able to break deployed mainnet contracts relying on eight known invariants. The 200M gas-limit bump was also removed, Sepolia remains live, and mainnet users need take no action while Chainsecurity, Dedaub and EF audits run ahead of the November 19 ACDE.

· Updated November 17, 2026 · Zain Tran · 7 min read · 0 total views · 0 today

Categories: technologyEthereum

Futuristic editorial illustration of the Ethereum Glamsterdam upgrade postponed on the Hoodi testnet with client release matrix

The epoch was burned into my calendar. Epoch 132,352. Slot 4,235,264. October 26, 2026, 17:42:48 UTC. That was the moment Hoodi, Ethereum's public staging ground, was supposed to swallow Glamsterdam whole: the Amsterdam execution fork, the Gloas consensus fork, and a gas limit jumping to 200M in the same breath.

Then the config changed.

On October 22, the Ethereum Foundation Protocol Cluster posted a short gut punch: the Hoodi activation "has been postponed." No new date. No replacement epoch. The scheduled jump of Hoodi's gas limit to 200M went with it. The post is careful to say this "rolls back the configuration for a future upgrade; it does not revert Hoodi's chain history." Clean language for: we unscheduled it, and nothing happened on-chain yet.

The reason is stated in plain sight. Core developers decided at ACDE #248 to delay specifically to investigate "interactions of EIP-8037 with mainnet smart contracts."

Key Metrics at a Glance

Metric Value
Postponed activation Hoodi epoch 132,352 / slot 4,235,264 (was 2026-10-26 17:42:48 UTC)
Gas-limit change 200M increase removed along with the activation
Trigger EIP-8037 state-creation gas repricing vs. deployed mainnet contracts
Invariants broken by EIP-8037/8038 8 (per EF public invariants repo)
External reports of breakable contracts 4 (as of the Nov 9 EF update)
Auditors engaged Chainsecurity, Dedaub, plus EF internal audits
Sepolia status Glamsterdam live since 2026-10-06 (epoch 353,024)
Mainnet activation Not scheduled; no action required from ETH holders
Next decision gate ACDE Nov 19, 2026 — no decision pre-committed

What the Delay Actually Removes

Read the post closely and notice what got pulled: not just the fork, but the 200M gas-limit schedule riding at the same epoch. That is the scaling promise — the whole reason anyone tolerates a hard fork — being explicitly unscheduled.

This is a config rollback, not a chain reorg. Hoodi keeps its history. But any operator who hardcoded the old activation time into their infrastructure now runs a client config that disagrees with the network config. That is exactly the mismatch that turns into a missed activation and a very long night.

The EIP-8037 Problem in Plain English

EIP-8037 raises, and separately meters, the gas cost of creating state: new accounts, new storage slots, deployed bytecode. EIP-8038 raises the cost of touching existing state: SSTORE, SLOAD, cold account access. State-operation gas prices were last repriced in the Berlin fork, back in 2021. State has grown a lot since then.

The EF's own August 24 post admitted the obvious consequence: replaying historical mainnet transactions under the new schedule surfaces a small set of contracts that rely on assumptions the new schedule shifts. The failure modes are specific and ugly — Solidity's transfer/send 2,300-gas stipend, hardcoded gas values in calls, logic that branches on gasleft(), and presigned transactions with fixed gas limits. "Potentially broken" means the contract fails even with a substantially raised gas limit.

Chainsecurity found projects relying on eight invariants that EIP-8037/8038 break. The EF published a public invariants repository and told developers to self-check. Then on November 9 the EF updated the post: four outside reports have identified deployed contracts that could break. Each report is being reviewed and the affected protocols contacted.

Here is the line that should make every deployer nervous: "we can not cover all projects on Ethereum." The EF is openly saying its audit cannot reach your contract. The burden of checking has been pushed onto you.

Ethereum testnet dashboard showing a postponed network upgrade timeline

Fork Status vs. the Promises

Stage Status Epoch / date Gas limit
Sepolia Activated 2026-10-06, epoch 353,024 testnet schedule
Hoodi Postponed was 2026-10-26, epoch 132,352 200M increase removed
Mainnet TBD not scheduled TBD

Glamsterdam itself combines the Amsterdam execution-layer changes with the Gloas consensus-layer changes. Its headliners are enshrined proposer-builder separation (EIP-7732) and block-level access lists (EIP-7928), plus the repricing pair at the center of this delay.

Client Releases Are Not the Schedule

The EF post lists Hoodi-compatible client releases — consensus layer: Grandine v3.0.0-rc.2, Lighthouse v8.3.2, Lodestar v1.50.1, Nimbus v26.10.2, Prysm v7.3.1, Teku 26.10.2; execution layer: Besu 26.10.1, Ethrex v30.0.1, Erigon v3.8.1, go-ethereum v1.17.10, Nethermind 2.2.1, Reth v2.8.1.

But the same post warns: "Do not treat the client release tables in the earlier announcement as confirmation that those versions incorporate this postponement." Translation: a release existing is not a release that knows the fork is off. Read the release notes, not the version number.

Upgrade Safety Margin Score

Proprietary framework: Safety Margin = (Invariant awareness x 0.35) + (Audit coverage x 0.35) + (Operator clarity x 0.30), each 0-10. Current Glamsterdam state: invariant awareness is high (8 published invariants, public repo) = 8; audit coverage is explicitly incomplete ("we can not cover all projects") with 4 confirmed breakable contracts and pending Chainsecurity/Dedaub reports = 4; operator clarity is decent (clear post, but release-table caveat) = 6. Weighted score: 0.35x8 + 0.35x4 + 0.30x6 = 2.8 + 1.4 + 1.8 = 6.0/10. The delay is the right call — but the pre-fork process had a hole the size of mainnet, and the score only reaches passing because the fork got pulled.

What You Should Actually Do

Decision flow:

What do you operate?
├── Hoodi node / validator
│     -> Update BOTH execution AND consensus clients to releases whose
│        notes explicitly remove the Oct 26 Hoodi activation
│     -> Never trust a version table over release notes
├── Deployed L1 smart contracts
│     -> Check address at ethereum.github.io/repricing-impact/
│     -> Test on Platåberget (glam-devnet-8)
│     -> Audit for: 8 broken invariants, 2300-gas stipend, gasleft()
│        branches, presigned-tx fixed gas limits
├── Wallet / RPC / indexer / gas estimator
│     -> Update eth_estimateGas handling for new cost rules
│     -> Cached gas constants will underestimate and fail txs
└── Regular ETH holder
      -> Nothing. Mainnet untouched. Do not plan around Hoodi.

Execution and consensus client release matrix visualization

The Accountability Layer

Someone at ACDE looked at EIP-8037, looked at mainnet, and said stop. That is the process working. But notice who actually finds the breakage: Chainsecurity during testing, then four volunteer reports, then the EF publicly asking developers to check their own contracts. The safety net for deployed, immutable contracts is largely self-service.

The November 6 post even concedes that changes affecting deployed contracts "aren't getting the scrutiny they deserve" — an admission embedded in an announcement. They are considering whether ACD should wait for public audit findings before revising the specs or scheduling Hoodi. Prototypes of possible spec revisions are already underway.

The delay is honest. The process that almost shipped it is the story.

Security audit workflow and broken invariant tracking dashboard

What to Watch

Per the EF timeline: Nov 12 — ACDC consensus agenda (EIP-8037/8038 not on it); Nov 13 — Dedaub and ChainSecurity reports delivered and published, engagements end; Nov 16 — ACDT findings presentation with open Q&A; Nov 19 — ACDE discussion of findings and next steps, no decision pre-committed. If audit findings show the impact on deployed contracts is too widespread to accept, ACD could revise the EIP-8037/8038 specifications before any new Hoodi date is set. Hegotá work is unaffected and continues in parallel.

TL;DR

  • What: Glamsterdam's Hoodi activation (epoch 132,352, was 2026-10-26 17:42:48 UTC) and its 200M gas-limit bump were postponed with no new date
  • Why: EIP-8037/8038 gas repricing can break deployed mainnet contracts riding on 8 known invariants; 4 breakable contracts have been reported
  • Who it hits: Hoodi operators must update clients; L1 developers must self-audit — the EF openly says its audit cannot cover all contracts; mainnet users are unaffected
  • Edge: Release tables are not the schedule — only client release notes that confirm the postponement count
  • Watch: Nov 13-19 audit publications and ACDE Nov 19; possible EIP-8037/8038 spec revisions before any new Hoodi date

Sources


Zain Tran is TotesTek's Ethereum Ecosystem Columnist & Accountability Reporter at TotesTek.