I tried this trick to see if it would work and I got some cool results.
I was rendering an animation where I have three computers with double GPU’s. On each of the machines I was rendering a frame at about 2:03 using 256x256 squares. So I opened another instance of blender on one of the computers and set it to work on the same animation but I had it render using a fixed number of threads at 6 threads of the 12 available. Then started it. The first instance slowed to about 2:08 per frame but the second instance is adding an additional frame every 9:48 so now I am making roughly six frames in 10 minutes instead of 5 frames. That is a 20% increase if I set it up on all three computers!
Just thought I would share and see if anyone else has experimented with this.