As a normal user your wallet could do that for you automatically.
The complexity could be wrapped through different adresses for example.
thats my main problem with LN or this extension block thing.
i feel you haven't covered all the bases and the user experience will be severely affected.
oh its simple, you just timelock your BTC into the Xblock and then you sign a raw TX, make sure you flip the bits because of big-endian bit ordering of course, and then you can withdraw from mtgox!
SOON
TM this will be all automatic.
good way to have everyone DASH the F out of here.
and the debate is truly silly when you realize 1MB is NOT AN OPTION, even with LN we'll NEED bigger blocks
just do it already.
Nah you are jumping one step ahead. This is only a raw idea.
If you'd have support from miners and would SF the logic into main blocks, you don't have to lock the bitcoins.
So how would this work without locking bitcoins:
One simple solution would be to hold them in a special address on the main utxo and the user adresses on the extension utxo.
All miners still mine the main blocks.
EC miners additionally mine ec blocks. Segwit miners also mine segwit blocks.
A lot of miners will mine both, because they want to have the fees.
This would require 51% of the miners of an extension block have to be honest, because only they know the rules.
So having 51% of all miners would be advisable.
When we compare this to a chain split scenario this is not worse, because there we also need 51% of the complete hashing power, to be safe against reorgs.
Sidenote:
I think a reorg on the main chain would only induce the same reorg to the extended blocks, because they are completely synchronized, if the logic of utxo selection for the downgrades is deterministic (think replay attacks) - this is where extension blocks differ from sidechains, where reorgs on the main chain are a huge problem.
(someone invent the next level of forum technology faster!) 
. 
