Overhead from inserting only one part of a complex model as a block in another file

If we insert a file as a block, but we only need an object that lives in two layers (and hide the others), that we then replicate 10000 times, is it efficient ?

There seems to be a bit of overhead from the hidden layers. For example Rhino almost hangs if done this way as opposed to isolating that object and save it to yet a third file on its own.