I have a mixed pair of GPUs: gtx670 and gtx770 with both enabled for use with BIONC on a win10x64 system. The system is core 2 quad so only 4 CPUs available. The GPUs both have 2gb and the same architecture so no large mismatch on the graphics boards.

They have been running a pair of Einstein tasks each using 0.5 for GPU and 1.0 for CPU. While each task ran slower, the fact they were running simultaneously made up for the difference in speed.

I allowed one PrimeGrid to run as a test. I expected that the board it would run on would have its 2 Einstein tasks bumped off into suspension. Not only were they bumped off, but the one of the 2 Einstein tasks on the other board was also suspended. Thus I had only 2 active tasks, 1 on each GPU and 2 unused CPU cores.

This should not have happened. Why was one of the Einstein tasks suspended from that other graphics board?

[EDIT]
I am running BOINC 7.8.3. Also want to mention that I did not suspend any of the Einstein tasks, I simply enabled "allow more work" on the PrimeGrid and a total of 6 WUs downloaded before I could revert back to "no more work". However, only 1 of the 6 WUs started.

I expected that the board it would run on would have its 2 Einstein tasks bumped off into suspension. Not only were they bumped off, but the one of the 2 Einstein tasks on the other board was also suspended. Thus I had only 2 active tasks, 1 on each GPU and 2 unused CPU cores.

This should not have happened.

Oh, I think it should, but "this" is not what you think it is.

Why was one of the Einstein tasks suspended from that other graphics board?

BOINC is clearing the other GPU for another Prime task. There's two individual Einstein tasks to suspend and restrictions to the timing to follow. Just watch for a while, the last Einstein task will be suspended soon.
This will continue to happen unless you can make Prime run on 0.5 GPU.