If your argument is that we publish the full transaction minus the public key and signatures, just committing to it, and then revealing that later (which means an attacker can't modify the transaction in advance in a way that produces a valid transaction);
Allowing expiration retains insecurity, while allowing expiration makes it a trivial DoS target.
Anybody can flood the miners with invalid transaction commitments. No miner can ever prune invalid commitments until a valid transaction is finalized which conflicts with the invalid commitments. You can't even rate limit it safely.
Like I said in the other thread, this is unreasonable. It's much more practical with simple hash commitment that you can "fold away" in a Merkle tree hash and which you don't need to validate until the full transaction is published.