Thanks for all the great observations!
After looking it over yet again, I think that the full-flow fuge as shown in sumpa.jpg would work fine, with the only issue being the high flow rate.
Adding the bypass loop as in sumpb will not provide the easy flow solution I hoped for. I agree with you guys in that it will introduce problems, with the fuge once again possibly having little or no flow. The suggestion of a check valve is a good one, but I now think my bypass idea was flawed. BTW, the bypass loop is shown underneath for clarity only. It would indeed have been level with the other plumbing. It's a moot point though, since I believe I now have a much better solution.
Combining some of our best ideas, I came up with drawing C which is attached. This moves the fuge back to the right, and feeds it using a tee and valve off of the return pump. I like this in that it doesn't require an additional pump. If, however, the return flow to the main tank is reduced too much, I could just then add a small pump to feed the fuge instead of tapping the main return flow.
The fuge outflows to the return box via gravity, through a standpipe. I'd much rather use a standpipe and keep all bulkheads at the same height on each box. Doing so provides consistency and redundancy, since boxes 1 and 3 will be identical, making it simple if I ever need to swap or reconfigure them. I won't run the fuge much deeper than the rest of the sump, since in the event of power failure I want all 3 containers quickly catching the overflow from the main tank. However, I can play around with that easily, since the depth of the fuge is varied merely by adjusting the length of the standpipe.
Let me know what you think of the latest incarnation.
Also, do any of you have the answer to what's the proper water depth for a TF1000?
Thanks again!