Ethereum EIP-7778 would count block work before gas refunds

A review deadline for block accounting

EIP-7778 lists November 1, 2026, as its Last Call deadline in the official document. Created on October 1, 2024, by Ben Adams and Toni Wahrstätter, this Ethereum proposal still shows Last Call when checked. It concerns block capacity accounting, with a central distinction: lowering a user’s bill should not erase work the network actually performed. Reaching this deadline establishes neither final adoption nor activation.

Gas measures and constrains resources consumed by operations. Some actions qualify for refunds, notably when an instruction clears a storage slot to zero. The current mechanism described by the proposal subtracts those refunds both from the user’s transaction cost and from the counter used to assess block fullness. The authors argue that this double subtraction can admit more computation than the limit is intended to allow.

Keep the refund and count gross workload

The text cites historical block 20,878,522 to illustrate the gap. Net usage is 28.5 million gas, with an additional 4.01 million in refunds, bringing gross usage to 32.51 million. In the authors’ example, that is 2.51 million above the block limit. This historical case is not an incident announced today and does not establish that every block exceeds its limit similarly.

The proposed solution separates the two accounts. Users’ transaction costs would continue to incorporate eligible refunds. Gas counted toward the block limit, however, would no longer subtract them. The formula also incorporates EIP-7623’s calldata floor by taking the relevant maximum. The proposal therefore does not abolish incentives to clean up state; it would prevent a financial rebate from being interpreted as computation that never occurred.

Block producers would need selection changes

Not every discount would be treated as an artificial refund of work. The document retains reductions reflecting genuinely lower resource usage, such as certain warm accesses or restoring an original value. That distinction matters: the model should constrain resources without adding work that did not occur to the block counter. Proposed tests compare storage-clearing transactions and blocks close to their limits.

The proposal is incompatible with previous rules and would require a hard fork. Block producers would need to update transaction-selection algorithms, while the model preserves individual refund economics. The authors see this as a way to constrain computational load and certain denial-of-service risks more effectively. The current development is the November 1 review milestone, rather than the removal of refunds or protection already deployed on the network.