well, if there are several versions , then anyhow the data has to kept somewhere; for different version you may have to add an attribute,a column, to identify the data version ; have that column included in clustered key.
but the point is that who is going to own the data.
Does this version process has to be live ? In a live environment you can't move the records, when there is a lot,and a user is on the front.
If it's a request process then you can have job running for each request.
To save the back end loader time ,If it's millions of row , then check if SCD works for your business environment; this might save a lot , if it suits your business.and for newer versions you may have fewer data to go forth and may be a lot less to roll back.
Archive versions do not need to be live. and I agree with your concern in regarding to moving records while user is connected. I will have to verify and see how the current process deals with the loader processing while user is on front ends. I bet the loader is normally scheduled to run at night time. So to minimize that risk, if user elects to rollback, the rollback will be to scheduled to run on another time, but that has another issue of its own if bad feeds are received, then the rollback may need to be done immediately.
I didnt understand what you mean by SCD, could you explain? I am terrible with Acronyms..
Regards,