Yeah that's what I ment....I thought it was pretty clear

It's something that can work in many cases to speed up renders, not all the time...
I'm not sure I understand the problems you think you would have doing something like this?
In Erics case for example, I think what I would have done (and maybe that IS what he did, and I'm confused), is instead of rendering 10 different interior renders with only the chair changing in each of them, I would:
- render the interior once with no chair, from the final point of view
- render one 360 view of the room, placing the camera where the chair(s) would be placed
- in another scene, create a plane to act as a shadow catcher, place the chair on the plane, and use the 360 as an IBL map. Repeat 10 renders like this changing the chair in each one. These renders would be super fast, compared to doing a render of each chair in the full interior scene, 10 times.
- composite each chair render into the first interior render, using the shadow channel to have realistic shadows.
So approximate time saved:
1x interior render to SL 18 = 5 hours
1x 360 HDR of that interior to SL 15 (I'm guessing you can denoise it a lot and slightly blur it, since it will only be used as an IBL map) = 2 hours, probably less.
10x chair renders to SL 16 = 20 min each (yes it would be that fast, perhaps even faster compared to a closed interior render) = 3.3 hours
TOTAL = 10.3 hours
compared to:
10x interior renders of each chair to SL 18 = 50 hours