Ethereum has exactly three contracts whose runtime starts with 0xEF. An August 28 update to Draft EIP-8298 says as much, but the old EOF EIP that names the three attaches the nine-byte farewell to the wrong address.
EIP-7610 was declined for Glamsterdam last week. It would have made CREATE and CREATE2 revert when the destination already had non-empty storage. The live Mainnet edge case is a small fossil: 28 accounts with storage, nonce zero and no code.
Those accounts do not look dead when you inspect their transaction history. Xatu has 100 direct transactions into 22 of them, and all 100 receipts say success. Ninety carried calldata. Twenty-five used selectors that resolve to withdraw or refund functions, while another 11 looked like attempts to destroy the contract that was not there.
At first glance, Mainnet made 122,453 positive-value transfers from an address to itself in 14 days. That is a silly number. It is also wrong for the question draft EIP-8358 is trying to answer.
EIP-5920 proposes a PAY opcode that moves ETH without running the recipient's code. It is Draft, and a pull request proposing it for inclusion merged on August 20. The pitch sounds like a cleaner, cheaper value transfer.
Mainnet has a better explanation. From July 2 through July 13, the Prism contract made 104,998 empty-calldata ETH calls to Uniswap v4's PoolManager. PoolManager rejected every one. Prism caught almost all of those failures and booked 0.813726 ETH as pending credit instead, which is exactly the branch PAY would skip.
I thought the missing corner of Ethereum's storage-lifecycle table would be ordinary transaction-local scratch space. It wasn't. Across August 5-18, 166,386 contract-slot cycles started at zero, became nonzero, and returned to zero before the block ended. USDC and USDT alone made 44,355 of them.
EIP-8374 looks like bookkeeping: stop deleting warm-access entries when an inner call reverts. Across 14 complete mainnet days, that one rule would have turned at least 169,495 later account accesses from cold to warm. The funny part is where they landed. 136,127 of them, or 80.3%, were calls into the same USDC FiatTokenV2_2 implementation.
This is not a chain-wide gas miracle. The verified address-only saving is 423,737,500 gas, just 0.0139% of the gas used in the window. It does expose a real execution shape hiding behind proxy calls: an inner frame touches an implementation and reverts, then a surviving frame reaches the same implementation and pays the cold surcharge again.
EIP-7709 came back from the dead last week. The proposal would charge BLOCKHASH for the storage reads it now represents, which sent me looking for contracts that call it hard enough to care. One SHIB airdrop did: 374 transactions into a single contract produced 149,600 calls, or 58.4% of every observed BLOCKHASH execution in a 14-day trace window.
The gas story is stranger than the count. Each airdrop read the same two ancestors 200 times, so most of the proposed storage charges would be warm. Even the cheapest all-warm backcast still pushes 327 of the 374 historical transactions past their original gas limits before their refund becomes usable.
Solidity's Panic(0x11) looks terminal. On mainnet it often isn't: across 14 complete UTC days, 5,333 of 5,772 transactions containing an overflow, underflow, or division-by-zero panic still ended green. Every panic happened in an internal child call.
A Hegotá proposal would remove a precompile that contracts called an observed 4.07 million times in twelve days. That sounds reckless until you look at the bill: a first-touch upper-bound backcast of Draft EIP-7666 adds up to 0.44% to the gas used by the affected transactions.
Ethereum's green receipt is a blunt answer to a messy question. Across 14 complete days, 288,119 transactions returned success even though at least one child call errored. Those transactions contained 1,225,260 failed child calls between them, or roughly one mixed result for every 115 successful transactions.