ckETH’s Sweeper Refactor Turns Deposit Consolidation Into a Signed, Replay-Safe Pipeline

The newest ckETH development is about settlement discipline rather than a new token. Between August 27 and 29, 2026, contributors merged a cluster of changes into the Internet Computer repository that reshapes how the ckETH minter consolidates automatic deposits on Ethereum.

The central change is architectural: a sweep request now identifies the token and the individual deposits it will move, while the transaction’s nonce, gas price, value, and call data are derived later by the pipeline. That separation gives the minter a clearer boundary between accounting intent and Ethereum transaction construction.

The refactor also closes a serious queue-consistency gap. Previously, batching could inspect deposits that had already been selected by another sweep, allowing the same balance to be offered to two transactions. The new design records which sweep holds each queued deposit and skips entries already in flight. Entries remain queued until the sweep settles, so a failed transaction can be represented and retried without losing ownership information.

Security checks are explicit. A deposit is excluded from a sweep if it lacks either the attestation connecting the deposit address to the credited ICP account or the authorization needed to delegate execution to the sweeper. The batch is also organized one sweep per token, reducing ambiguity when multiple ERC-20 assets are processed in the same cycle.

Gas accounting receives a corresponding correction. The old fixed 100,000-gas assumption could underprice batches that touched several addresses and installed EIP-7702 authorizations. The new limit grows with the number of addresses visited, and the maximum transaction fee is calculated from that same estimate. In practical terms, the minter is less likely to create a transaction whose own gas cap makes inclusion impossible.

For builders, the important lesson is that Chain Fusion reliability is not only about RPC quorum or threshold signatures. It is also about durable state machines: every selected deposit needs an owner, every authorization needs a proof, and every retry needs a recorded failure path. The ckETH documentation describes the surrounding model: Ethereum deposits are discovered through helper-contract events and EVM RPC access, while withdrawals are signed and submitted by the minter.

The caveat is important: these are upstream implementation commits, not a published performance benchmark or confirmed user-facing release; their final behavior may change before deployment. Developers integrating with ckETH should therefore rely on the released Candid interface and current documentation, and should re-check the minter’s live interface before depending on unreleased sweep events.

If the work reaches production, its main impact may be invisible to users: fewer duplicate sweep opportunities, more predictable retries, and transaction fees that better reflect batch size. That is precisely the kind of low-level accounting hygiene that makes an ICP-hosted cross-chain asset safer to compose.