Multisig based on public key hashes instead of public keys

Is it possible? And if so, why don't we use public key hashes instead of public keys? For example, let's assume we have first two private keys:

Code:

alicePrivKey=0000000000000000000000000000000000000000000000000000000000000001
bobPrivKey=0000000000000000000000000000000000000000000000000000000000000002

Then, each party can compute its public key:

Code:

alicePubKey=0279be667ef9dcbbac55a06295ce870b07029bfcdb2dce28d959f2815b16f81798
bobPubKey=02c6047f9441ed7d6d3045406e95c07cd85c778e4b8cef3ca7abac09b95c709ee5

And then, each party can compute its address:

Code:

aliceAddress=bc1qw508d6qejxtdg4y5r3zarvary0c5xw7kv8f3t4
bobAddress=bc1qq6hag67dl53wl99vzg42z8eyzfz2xlkvxechjp

Sending to address is enough to get it recorded on-chain. If coins were never moved, then it is impossible to know the public key. But it may be possible to create a multisig that is based on such address. For example, we can start from getting hashes for each of them:

Code:

aliceHash=751e76e8199196d454941c45d1b3a323f1433bd6
bobHash=06afd46bcdfd22ef94ac122aa11f241244a37ecc

And then, we can create 1-of-2 multisig, when revealing only one public key is needed to move the coins:

Code:

OP_DUP OP_HASH160 OP_DUP 751e76e8199196d454941c45d1b3a323f1433bd6 OP_EQUAL OP_SWAP 06afd46bcdfd22ef94ac122aa11f241244a37ecc OP_EQUAL OP_ADD OP_VERIFY OP_CHECKSIG

As far as I know, it should work in theory, but I still cannot confirm it would work in practice. Both paths should be spendable. For Alice, it would be:

Code:

<sigAlice> <pubkeyAlice>
<sigAlice> <pubkeyAlice> <pubkeyAlice>
<sigAlice> <pubkeyAlice> <hashAlice>
<sigAlice> <pubkeyAlice> <hashAlice> <hashAlice>
<sigAlice> <pubkeyAlice> <hashAlice> <hashAlice> 751e76e8199196d454941c45d1b3a323f1433bd6
<sigAlice> <pubkeyAlice> <hashAlice> 1
<sigAlice> <pubkeyAlice> 1 <hashAlice>
<sigAlice> <pubkeyAlice> 1 <hashAlice> 06afd46bcdfd22ef94ac122aa11f241244a37ecc
<sigAlice> <pubkeyAlice> 1 0
<sigAlice> <pubkeyAlice> 1
<sigAlice> <pubkeyAlice>
1

And for Bob, it should left "0 1" instead of "1 0" on the stack, so after addition, OP_VERIFY will succeed by consuming OP_TRUE. I am still trying to confirm that on regtest, but I think it should be technically possible to use public key hashes to make a multisig, instead of using public keys.