In the UI the user sees whatever the add-in which performed the background open or queries the document opened by another add-in wants the user to see by the add-in user interface. The data can only be accessed via the api though.
Not much - opening time should be basically the same, excepting the time to ‘draw’ the view which the UI would require. Memory and compute load would be equal as well excepting that ‘draw’ aspect, which for most projects is minimal. The difference is that in one method the user can interact with the data via the UI and in the other method they cannot.