Did you try running the same encode twice, once each way, to see if the encode is actually slowing down? If the encode rate is actually dropping, you have a problem. If the system is doing the same work with less power and noise when minimized, you have the opposite of a problem.
That was a very good idea. I was looking at the fastest cores 5,6,7,15 and just noticed the speed drop...several hundred Mhzs.
But when I switched back and forth foreground/background I was watching the graph and saw that yes the top core speeds dropped BUT most of the slower cores went up by 100 Mhz +/-.
So the end results was an 8 seconds faster encode (4:22 vs 4:30) in the background then the foreground! Nice surprise.
EDIT...VideoReDo also gains fps by being minimized...285fps vs 281fps. Consistent results.