The difference is that users would soon realize that they’re not getting the service (not getting an NFT or not being able to withdraw - although that’s trickier until withdrawals are enabled).
With EXTRACTGAS contracts will be able to provide a service, while the user pays a premium unknowingly. Something like this - a new dex targeting Uniswap users, so it airdrops tokens to all existing Uniswap projects with a short expiration date to induce FOMO. Users claim the token, while unknowingly paying for these tokens via EXTRACTGAS. A token presale disguised as an airdrop. Users will see that they’re getting the claimed token, not realizing that they paid for them.
The way to protect the user against such scams will probably be through EOA wallet software like Metamask, which should simulate transactions and clearly warn the user that this transaction is going to pay X to contract Y. With proper UI it could work. To help Metamask do that, maybe we should add an extractedGas field to the return value of eth_call or add a new eth_estimateGasWithExtraction call which tells the caller, not just how much gas is used, but also how much gas is extracted by contracts using EXTRACTGAS.
The ultimate purpose of account abstraction is to replace EOAs. ERC 4337 is just the beginning and requires creating a contract wallet, but the endgame is that at some point (in a distant future) there will be a hard fork that replaces all EOAs with a proxy contract that points at a precompile wallet emulating an EOA. Users will then be able to upgrade their proxy in-place to point to more sophisticated wallets.
In the short term I see the value of such opcodes, but in the long term it might fall in the “enshrining-EOAs-even-further” category.