Reporting window: September 7, 2026, 20:16:35 through September 14, 2026, 20:16:35 WITA (Asia/Makassar).

Hegotá’s scope became more selective this week. ACDE formally made EIP-8141: Frame Transaction a headliner while leaving its specification open to revision. Meanwhile, Glamsterdam testing exposed another concrete interoperability gap: the execution-specs team merged tests for missing system-phase access-list entries. My reading is that implementation capacity now sets the agenda: Hegotá candidates need a stronger case for priority, while Glamsterdam needs evidence that clients agree on edge cases.

Compiler demand becomes a scoping test

EIP-7979: Call and Return Opcodes for the EVM survived an immediate deferral decision after Jochem Brouwer challenged client developers’ ability to judge EVM-user priorities and Greg Colvin defended the proposal. Ansgar Dietrichs postponed the decision so compiler teams could participate.

That reprieve did not extend to EIP-7923: Linear, Page-Based Memory Costing: the call deferred it from Hegotá. The distinction matters. Technical merit alone did not secure scope; the next useful signal is whether compiler teams make a concrete adoption case that changes client priorities.

Glamsterdam gets a more explicit rehearsal

EthPandaOps’ devnet-11 plan schedules genesis for September 14 and the Gloas transition for September 16 at 20:00 WITA, with explicit specification and fixture requirements for client images. It describes a normal fork-transition rehearsal; adversarial testing remains on devnet-8. This is a published plan, not confirmation that genesis or the transition succeeded.

Scope discipline also held: ACDE declined the late addition of EIP-8253: Bump nonce of zero-nonce storage accounts, after Toni Wahrstätter raised testing and BAL-interaction concerns.

Sepolia’s October 6 activation remained a proposal with Hoodi still undated. The confirmation discussion was scheduled after this reporting cutoff. Watch the rehearsal outcome and that decision before treating the date as settled.

Access-list correctness reaches beyond transactions

The new Block-Level Access Lists tests address omissions that can leave both state root and gas unchanged. Author ilitteri reported that ethrex’s default parallel import path accepted missing system-call reads that Geth, Reth and Erigon rejected. The September 8 merge establishes regression coverage; it does not establish universal client adoption.

A separate September 9 specification amendment removed the prescribed remaining-gas feasibility bound for storage reads. System calls do not pay block gas, so that bound could reject valid blocks. Together, these changes sharpen the requirement: clients must catch incomplete BALs without imposing invalid assumptions on system-phase accesses.

More capacity does not settle the fee question

In From 60M to 200M, misilva73 published twelve demand scenarios on September 9. Nine settled near target utilization; three remained below it despite negligible base fees.

The model’s limitations are consequential: it fixes the transaction mix and does not re-execute sampled transactions against evolving state. Treat this as a guide to what to measure after repricing, not a fee forecast or an activation decision.

Sources and further reading

All sources fetched 2026-09-14; no archive snapshots obtained.