After the last seizures of "wallets" from authorities, I have been wondering about how to prevent them from extracting the passphrase through chemicals and/or torture.

I know there are already a nice set of solutions, which include at least:
- Multi-signing
- Hidden cold wallets
- Pre-signed transaction "bombs" that an accomplice can trigger on a "jailed man switch".

But I recently watched the latest BBC Sherlock series (which I found surprisingly good, btw), and in one of the episodes, [Spoiler]Sherlock gets his hands on an encrypted device. Problem is that device can react to 2 keys : one that will unlock and let you access the data, one that will destroy the data. So even if they torture the device owner, he still has the possibility to give out the "self-destruct" key. [/Spoiler]

I liked that idea.
So I understand this is technically impossible to acheive today, because it would require some encrypted code to be executed within the wallet itself, because if in a client (like a walletpassphrase method), they can just design their own without any risk to trigger the self-destruct actions.
I also understand they would probably not be connected to the internet while trying out the passphrase.
Just corrupting the wallet file would obviously not be enough either, as it's just a copy anyway and the real data is out there in the blockchain.

Any idea about that?