Exactly. If a solution is not understandable for users with average IT expertise, then it will never be understandable for anyone with even less IT knowledge. And typically owners of large mining farms and exchanges do not have time to do those learning, so they tends to select the solution that they can understand or listen to people they like. This will turn the decision making into politics, and who are good at lobbying and PR will push in their changes. And this is not people would like to see in bitcoin. So, the knowledge gap of different participants decided that you really can't reach a wide consensus upon a radical or complex solution, XT's failure already proved that
Understanding can be of different levels: conceptual, algorithmic, implementational... I bet most people don't quite grasp how Bitcoin's Script stack machine is implemented, though it doesn't prevent them from utilising it, if they know conceptually at least. What's enough for most people is that a particular component has been peer-reviewed thoroughly to prove it's safe to use it.
I still don't really understand how that can be implemented as a soft fork. Softfork means backward compatible, when the upgraded SW clients broadcast new blocks through out the network, how come the original core client can accept such kind of strange block which does not contain signature data?
There are two modifications to be made for it to be soft-fork compatible:
1) SW outputs are made
anyone can spend, so that for older clients it won't matter how they are spent, the scriptSig will be empty.
2) The merkle tree root of SW data hashes is stored in the coinbase.