Reporting window: September 14, 2026, 20:16:35 through September 16, 2026, 22:30:00 WITA (Asia/Makassar). This edition covers the shortened interval since the preceding weekly cutoff.

Glamsterdam’s Sepolia schedule moved into the merged coordination record: Nixo merged Parithosh’s activation-timing proposal on September 15. Alongside that step, maintainers merged FOCIL harness corrections and a fork-choice test-vector ordering fix. My reading is that the immediate priority is making readiness evidence trustworthy: an agreed timing record matters, but so does knowing which red tests actually implicate clients.

A dated Sepolia, with limits on what the date proves

The merged timing proposal specifies October 6 at 21:53:36 WITA, epoch 353024, for Sepolia; Hoodi and mainnet remain TBD. This is a change in the coordination record, not evidence that released clients already carry the configuration or that a devnet transition succeeded. The next useful signals are matching client releases and a documented clean transition before that public-testnet activation.

FOCIL’s red board needs a better test oracle

ilitteri’s simulator fix merged into devnets/focil/0 on September 15. His diagnosis: the harness demanded a non-null inclusionListSatisfied for invalid payloads, although the API requires null. The PR reports 5,489 affected fixtures across four clients, with a selected 49-case rerun improving from zero passes to 49. That is evidence of a corrected assertion and a bounded rerun, not a clean full devnet.

A companion patch, also merged September 15, repairs fixture generation: stale blob-gas headers and an unrealistically low block-gas budget accounted for another 127 bad expectations. Its author says regenerating and releasing the fixtures is still necessary. Watch the next fixture release and full Hive run; the merge alone cannot establish their outcome.

The consensus side has a similar lesson. On September 16, mkalinin merged eserilev’s correction to FCR attestation-step ordering: exported steps placed attestations before the slot tick, while the generator applied them afterward. Lighthouse rejected that replay as future-slot input; Prysm’s reviewer reported an existing workaround. Aligning the vectors removes a source of misleading cross-client comparisons.

Gas accounting closes two specification gaps

Two September 14 merges sharpened EIP-8037: State Creation Gas Cost Increase. chfast’s rollback clarification prevents pre-execution state charges from being returned when the paid-for state survives a top-level revert. The PR explicitly says this documents existing reference-implementation behavior. His separate total transaction-gas cap bounds tx.gas at 4,294,967,295, alongside the separate execution-gas limit. These are merged specification changes; they do not establish adoption by every client.

For EIP-8038: State-access gas cost update, Maria Silva’s approval preceded the September 14 merge of flcl42’s warm-beneficiary SELFDESTRUCT clarification. Warm beneficiaries retain zero access charge; the same clarification removes the misleading 100-gas implication without changing the opcode’s base cost or conditional write and state-creation charges. The implementation question to watch is whether clients and fixtures agree on these exact boundaries, especially when execution reverts but earlier state changes survive.

Sources and further reading