The key thing here is that it requires a hard-fork in the Bitcoin network, and possibly destroy a lot of balances depending on the way the change is implemented.
Thinking a bit about SIGHASH_ANYPREVOUT, it would provide almost everything an "account" model would need. And SIGHASH_ANYPREVOUT is a softfork afaik.
What is an account model exactly? The big difference is which coins are transferred when you sign a transaction, and what the "state" of the network means.
In the UTXO model, the UTXO is "consumed" and a new UTXO is created, and the new UTXO is stored in the UTXO set. There is no other information needed.
In the account model, you will transfer any coins lying on that address (account), and full nodes process the transactions and store the balances of the addresses.
The main difference is the "transaction processing" step. In the account model, if you have more than one transaction involving a single account in a block, the exact order matters, due to possible replay attacks. Thus, there is a mechanism called "Nonce ordering", where a kind of "counter" is added to each transaction involving a single account.
In the UTXO model instead, an UTXO can only be consumed or not, no balance can be added or deducted. So you can process all transactions step by step and can parallelize that. Of course chains of zero-conf transactions are possible, but miners would include them in a block as long as all UTXOs are correctly spent.
Now what would happen with SIGHASH_ANYPREVOUT? Here there is a similar replay possible like in Ethereum, so there must be a transaction ordering
in the protocol using that feature (e.g. Eltoo) with additional constraints on the transaction. In Eltoo for example the sequence number is used in a similar way to the nonce ordering in Ethereum. The "state ordering" is thus achieved with smart contracting means, not natively.
But that replay risk is also what enables smart contracts to work a bit like in Ethereum and being able to send funds as a "reaction" to other events without knowing the exact UTXO (again, like in Eltoo), although other mechanisms are still necessary to achieve Turing completeness.
OP_CTV seems to be
able to simulate SIGHASH_ANYPREVOUT and enable also more contracts, so it would also enable "account-like behavior" on Bitcoin.
I wonder if this reasoning is correct and where there are still advantages of the "account" model which SIGHASH_ANYPREVOUT nor OP_CTV can provide.
