Hi Frank, welcome!
With referencing, we are definitely aware of the limitation you're hitting. Solutions have been discussed internally but I'm not able to give you a date on when those will be implemented.
As for simulating from origin, you can move into any position in world space over one frame, so long as you move the zSolver at the same time. All of Ziva's forces are computed relative to the solver position/orientation. This means moving them (character and zSolver) at the same time will negate the forces that would normally act on the character. I would advise doing this over 1 substep. (substeps are keyable)
From there a 10 frame transition from default pose in world space to animation start pose should be enough to run into your simulation safely. Because it's a physical simulation there isn't a way to shortcut this step unfortunately!
Most of the workflow for building fat is described in this webinar:
We ran into tech issues with the last part (probably the most relevant part to your question 🙂 )
For that part you take your fascia, duplicate it, and project the duplicate using Wrap3 (or other tools) to the skin mesh. Then you can easily bridge the border edges between the fascia layer, and the projected duplicate. Now you have a water tight fat mesh that you can simulate. You can just wrap the skin model to the alembic bake of the fat, without going back to the modeling department for approval, because you haven't changed the skin model.
You can also run a coupled "epidermis" when you run your fat pass. The epidermis is usually high enough resolution to capture fine wrinkles. It's coincident with the outer layer of the fat. You solve this as Ziva cloth at the same time as the fat. In this case, you would wrap the skin to bake of the cloth, as opposed to the fat.
If for some reason (I can't think of a good reason at this stage, other than a quirk of the production) you need the output of the simulation to match exactly the topology of the skin model, you could potentially make the epidermis a duplicate of the skin model. (with hands/feet/head deleted)
As for the last question -- it's possible to scale Ziva rigs. You just need to scale the bones, and the zSolver in lockstep. Everything else will come along for the ride.
Let me know if you have any questions about any of that! Hope I was able to address your queries.