They're essentially essay questions- come up with reasonable-enoguh assumptions and get on with business.
We don't do a -ton- of stuff like that, but I've found that if you can't estimate with any rationale, you're probably just not very good at whatever domain you're asked to deal with.
For example, 'what sort of architecture are we going to use to make this system', 'how many simultaneous users are we expecting', 'how many requests per second can we handle given this hardware', 'how much memory do we really need- if we quadrupled it, what would that get us', etc.
Also- if a person locks up when they're thrown "How would you move mt fuji", there are going to be a lot questions that I'd be concerned about getting answers out of you for...
For a Sr or higher person, I'd hope to get some reasonable answers.
Am I off?