Ethereum’s Goerli Testnet Finalized Shapella After a Slow Start

Ethereum’s Goerli testnet finalized the Shapella upgrade on March 15, 2023, completing the last planned public-network rehearsal before developers considered enabling validator withdrawals on mainnet. The result was consequential because Ethereum’s proof-of-stake system still allowed deposits but not withdrawals on its production network.

The fork was configured for Goerli epoch 162304, scheduled for 22:25:36 UTC on March 14. Finality came on March 15 UTC after a delay attributed to testnet validators that had not upgraded their software. An Ethereum client developer involved in the event reported that Goerli finalized 15 epochs after the fork point and said no client-software fault had been found; offline operators had to update and return.

That outcome was a pass with an operational warning, not proof that mainnet was ready. Goerli used valueless test ether, giving volunteer operators less reason to maintain nodes than validators securing economically valuable ETH. Its participation pattern could expose coordination failures, but it could not reproduce mainnet incentives or guarantee an uneventful production fork.

What Goerli tested

“Shapella” joined Shanghai, the execution-layer upgrade, with Capella, the consensus-layer upgrade. The two layers had to change together: the Beacon Chain determined which validator balances were eligible, while execution clients credited the designated Ethereum addresses.

The headline change was EIP-4895. It defined withdrawals as system-level operations carried inside execution payloads, not transactions submitted by recipients. Capella also supported full withdrawals for exited validators and partial withdrawals of eligible balances above a validator’s effective balance. Validators with older BLS withdrawal credentials first had to submit a one-time message assigning an execution-layer address.

The Ethereum Foundation’s March 8 announcement described Goerli as the last testnet upgrade before mainnet scheduling. It also warned that node operators running incompatible clients would remain on the pre-fork chain. The delayed finality on March 15 therefore tested software deployment and operator coordination alongside the withdrawal rules themselves.

A downstream incompatibility surfaced

The fork also exposed a concrete integration problem outside Ethereum’s core clients. Optimism’s status record said its Goerli nodes began halting when Shanghai changed the way the layer-one block hash was derived. Checks inside Optimism’s op-node recomputed that hash and began failing after the Ethereum testnet fork.

Optimism said its sequencer continued processing transactions, deposits and withdrawals safely while affected node operators applied a temporary trust-RPC flag. The status page marked that initial incident resolved at 08:05 UTC on March 15. This did not invalidate Ethereum’s upgrade, but it showed why a public rehearsal mattered: downstream software could depend on execution behavior changed by a fork even when the base protocol clients worked as designed.

What was still unknown on March 15

As of March 15, developers had not yet announced a mainnet activation date. Goerli had demonstrated that withdrawal-capable rules could activate and finalize on the broadest planned test network, but the slow start and Optimism incident left operational lessons to absorb. No event-day price claim is warranted here; the protocol records do not isolate Shapella’s effect from the banking turmoil and broader market forces surrounding cryptocurrency trading that week.

Later context

On March 28, the Ethereum Foundation described the Goerli transition as smooth and announced an April 12 mainnet target. That later decision clarifies the testing sequence, but it was not known on March 15. The event-day record ends with Goerli finalized, withdrawals being exercised on testnet, and mainnet scheduling still pending.