Is it possible to move a work unit to a different GPU?

[JustusLM]

I use my main system for folding on the two GPUs only (the CPU isn't worth it). They are identical Nvidia GTX 1080s.
Now, it happens again and again that I get two work units of varying complexity, and one completes in 1-2 hours and the other one in 5 or so.
If I am only away from my PC for 2 hours, I like to gamble, start folding and hope I get short units.
The good thing is, that as long as a longer unit lands on the secondary GPU's slot, I can still watch videos or even play some games that don't support Sli with hardly any performance impact.

Sometimes, however, it doesn't work out like that, and the "long unit" lands on slot one, and thus on GPU 0, meaning I can't use my PC for anything except browsing if I don't pause the work unit.
I've been trying to find a way to get this work unit onto the second GPU, I tried setting the slots up to usae the other GPU, but for some reason, that doesn't work.
In fact, even if I put in GPU indexes that aren't even present theoretically (like 2) they still use the same GPUs.

Looking into older threads, the only other method I could find is moving work unit data found in the ProgramData folder or in the AppData folder, depending on the client.
However, both of these apply to older clients, as neither of those folders exist on my computer (the "FAHClient" folder in AppData/Roaming, or /Local, or /LocalLow or the "work" folder in ProgramData/FAHClient).
I haven't found anything else on this yet, so I was wondering if anyone knows how to move a work unit to a different slot or force the slot to use a different GPU on a newer client.
Any help, even if it's just a guess, would be highly appreciated.

[bruce]

If you REMOVE the slow GPU, that WU should move to the only remaining slot (leaving two WUs enqueued). (That's assuming that FAH sees the GPUs as identical, like you said) When you re-add that GPU slot, it will download an new WU.

Note: this may cost you more PPD that it saves, depending on which of the two GPUs gets processed first and how long they both run -- since you are, in fact, intentionally delaying one of them.

I don't know of any way to re-separate them back into different slots.