I’ve been watching Ethereum’s upgrade cycle for years, and every time a testnet date gets locked in I half-expect a last-minute curveball. This week that curveball arrived in the form of EIP-8037. What looked like a straightforward path toward higher throughput on the Hoodi testnet suddenly paused, leaving node operators scrambling and developers rethinking assumptions about how gas behaves when calls fail. The October 26 activation is off the table, the planned jump to a 200 million gas limit is on hold, and the conversation has shifted from “when” to “how carefully.”
Why the Glamsterdam Pause Actually Matters
Glamsterdam was never just another fork name. It bundles enshrined proposer-builder separation with block-level access lists, two changes meant to make block production cleaner and more predictable. On paper the combination should help Ethereum process transactions more efficiently while reducing certain forms of MEV extraction. In practice, the execution-layer piece that drew the most scrutiny is EIP-8037, which separates the accounting of gas spent creating new state from gas spent on everything else. A companion proposal, EIP-8038, tweaks the cost of reading existing state. Together they alter the economics and the failure modes of external calls.
ChainSecurity published a technical assessment on October 13 that flagged a subtle but real problem. Certain contracts treat remaining gas as a signal. If an external call runs out of gas, the caller can look at how much gas is left and decide whether the failure was permanent or worth retrying. Under the new accounting rules a call that would have fully reverted might now leave enough gas for the calling contract to continue and record the failure as final. The security firm said it had already located thirteen live examples on mainnet. Responsible disclosure is underway, which is why the specific contracts have not been named.
A failed operation could leave enough gas for the calling contract to continue processing when the same transaction would previously have reverted.
That distinction sounds academic until you remember how many DeFi protocols batch operations, handle withdrawals, or implement retry logic around external calls. I’ve found that the contracts most at risk tend to be the ones that try to be clever with gas rather than the ones that simply let everything revert. In my experience, those clever patterns often exist for a reason—users expect a clean failure they can try again—yet the new gas rules can quietly break the expectation without any code change on the contract side.
What Hoodi Operators Need to Do Right Now
The Ethereum Foundation announced the postponement on October 22 after an All Core Developers Execution call. The original Hoodi activation was set for epoch 132,352, which lined up with 17:42:48 UTC on October 26. Both the Amsterdam execution-layer activation and the Gloas consensus-layer activation have been stripped from the revised configuration. Importantly, this does not rewind any blocks already produced on Hoodi; it simply removes the scheduled fork.
Node operators who had already updated to the previous release set must now move again. The Foundation listed Lodestar v1.50.1 for the consensus layer and pointed to Besu 26.10.1 and Nethermind 2.2.1 on the execution side. Ethrex is still in prerelease territory with v30.0.1-rc.2. Anyone running validators should check both the beacon node and the validator client, plus any external block-building setup they rely on. Christine D. Kim posted a direct PSA on X urging operators to revert or wait for client teams to ship versions that explicitly cancel the upgrade. Releases that still contain the old activation timestamp should not be treated as safe.
- Confirm your execution client no longer schedules the Amsterdam fork.
- Update the consensus client to a version that drops the Gloas activation.
- Verify both beacon and validator software if you are staking.
- Watch for final client releases that explicitly document the cancellation.
Sepolia already activated Glamsterdam on October 6 and has been running with the higher gas limit. That testnet is unaffected by the Hoodi decision. The two environments serve different purposes, and the fact that Sepolia continued does not guarantee Hoodi will follow the same path once the compatibility questions are resolved.
How EIP-8037 Changes Failure Handling
To understand why the security firm raised the alarm, it helps to walk through the gas accounting change without the jargon. Today, when a contract makes an external call and that call consumes all remaining gas, the entire transaction typically reverts. The calling contract never gets to execute code after the failed call. Under EIP-8037 the gas used to create new storage is tracked separately. A call that fails because it tried to write too much state might still leave the caller with usable gas. The caller can then mark the operation as permanently failed and move on, even though the original design assumed a full revert would give the user another chance.
Perhaps the most interesting aspect is how quietly this can break existing logic. No Solidity source needs to change. The EVM simply behaves differently around the boundary between “out of gas” and “out of state-creation gas.” Contracts that use patterns such as “if gasleft() > threshold then retry else record failure” can now take the else branch when they previously took the if branch. ChainSecurity noted that the affected patterns appear in batched transactions, certain withdrawal flows, and other multi-step operations that treat gas exhaustion as a soft signal rather than a hard stop.
| Scenario | Current Behavior | Under EIP-8037 |
| External call runs out of gas | Full transaction revert | Possible partial continuation |
| State-creation gas exhausted | Usually full revert | Caller may retain gas |
| Retry logic based on gasleft() | Often triggers retry | May record permanent failure |
| User expectation | Clean failure, try again | Risk of silent finalization |
I keep coming back to the thirteen real-world examples. Thirteen is not a huge number relative to the entire Ethereum mainnet, yet each one represents a contract that users already trust with value. If even a subset of those contracts can lose the ability to retry under the new rules, the upgrade has a compatibility surface that developers did not fully map before the Hoodi date was set. That is the practical reason the core developers chose to pause rather than push forward and hope for the best.
Timeline Shifts and What Remains Undecided
Before the postponement, developers had floated December 1, December 8, and December 9 as possible mainnet windows. The October 22 agenda even included a plan to pick a date the following week if Hoodi testing looked clean. None of those dates were ever confirmed, and the latest Foundation note does not replace them with new ones. The team must first finish examining how EIP-8037 interacts with existing mainnet contracts before another Hoodi activation can be scheduled.
Mainnet users and ETH holders do not need to do anything. Application developers have been told not to plan further Hoodi testing around the canceled October 26 slot. The gas-limit increase that was meant to ride along with the upgrade on Hoodi is also off the table for now. That increase was intended as an observation exercise; it does not reverse any history and can be re-introduced later once the compatibility work is done.
Looking ahead, the pause is a reminder that Ethereum’s multi-client, multi-testnet process exists precisely for moments like this. Sepolia has already given the community one data point on higher gas limits. Hoodi was supposed to provide a second, more production-like data point before mainnet. By stopping, the developers kept the second data point from becoming a source of unexpected mainnet behavior. Whether that decision ultimately shortens or lengthens the overall timeline depends on how quickly the affected contracts can be reviewed and, if necessary, patched or worked around.
Broader Implications for Developers and Protocols
Protocols that rely on precise failure semantics should start reviewing their external-call patterns now, even if the Hoodi date remains unknown. The risk is not limited to contracts that explicitly check gasleft(). Any code path that assumes an out-of-gas failure will always revert the entire transaction could behave differently once state-creation gas is accounted for separately. Curve Finance had already flagged that revised gas charges might affect certain DeFi applications; the ChainSecurity findings sharpen that warning.
One practical step is to add explicit gas-limit arguments to external calls rather than relying on the remaining gas of the parent transaction. Another is to treat any failure as potentially permanent and design the user flow accordingly, rather than hoping the EVM will always hand back a clean revert. These are not new recommendations, yet the EIP-8037 discussion makes them more urgent. I’ve found that teams who treat gas as a scarce resource rather than an infinite pool tend to surface these edge cases earlier in testing.
- Audit external calls that check remaining gas.
- Replace implicit gas inheritance with explicit limits where possible.
- Document expected failure modes for users and integrators.
- Test against both current and proposed gas accounting rules.
- Coordinate with the responsible-disclosure process if your contract is among the thirteen.
The enshrined proposer-builder separation and block-level access lists that sit alongside the gas changes are still valuable. They address real inefficiencies in how blocks are built and how state is accessed. The pause does not cancel those features; it simply refuses to ship them until the gas-accounting side effects are better understood. In that sense the delay is a feature of the process, not a failure of it.
Looking Past the Immediate Delay
Ethereum has postponed testnet activations before when compatibility questions arose. Each time the network has eventually moved forward with clearer rules. The current situation is no different in kind, even if the specific technical trigger is new. What feels different this cycle is the speed at which a security firm could surface concrete mainnet examples and the willingness of core developers to cancel a date rather than squeeze the investigation into the remaining days.
For anyone running a node, the immediate task is mechanical: update clients so they do not attempt the canceled fork. For anyone building applications, the task is analytical: understand whether your contracts treat gas exhaustion as a signal and, if they do, decide whether that signal still means what you think it means. The rest of the ecosystem can treat the postponement as a normal, if unwelcome, part of shipping a complex protocol upgrade. The December windows remain possible, but they are no longer the working assumption. The working assumption is that another Hoodi date will appear once the EIP-8037 questions have answers.
Until then, the most useful stance is the one the Foundation already took: keep mainnet stable, keep Sepolia running, and give the security review the time it needs. The Glamsterdam features are still coming. They are simply coming later, and with a clearer map of the contracts that might notice the difference.
