The result depends on how you import, from which starting format. In JSON-LD, you can select whether new elements should overwrite old elements, be added as copy, or not imported.
That aside, it is not a good idea to have all ecoinvent system models in one database, as they contain the same process datasets (with only some few different processes) but with a different modeling idea; mixing them in one of your models will make the model inconsistent. We have configured the databases so that the processes are not mixed in the linking algorithm, but when you yourself create the foreground model then you will really need to pay attention.
Hth, best wishes