I'm not sure how it can be "delegated" and not require trust in some form or fashion.
You didn't pay close attention to the phrases "
verifiable truth" and "
they make no discretionary choices of significance, persist the minimum state possible, are fungible with each other" I bolded in the generous excerpt that I quoted from my white paper draft.
If I delegate to you, but you have no choice in how you execute the delegated task, then I don't have to trust you.
I will not tell you the precise inventions of how I do that. Absolutely no hints will be given on that for now. Sorry.
On the surface it sounds a little like our old "hatcher" process, where transactions were delegated to a 3rd party of the clients choice. No trust was required as the output from a hatcher would either be correct, the minimum state as you put it, or not. If the latter, it would be apparent to everyone and that work would be rejected along with the transaction.
There were "issues" with this approach that I wasn't happy with and found myself running around in circles. So I dropped it, quite recently in-fact (past 6 months or so) and replaced it with a network wide solution we have now.