I agree. It would have been fun to test it on a properly designed board with traces and extra components that maintain the correct impedance and such. Fundamentally I think it could have worked for 100Mbit if designed properly. I believe one of the pages even stated having achieved decent results with 100Mbit.
I did the "best" I could with tape to keep the pairs as close together as possible. I am not too worried about untwisting too much, but more worried about the distance between the parallel wires in each pair.
The circuit is made out of 2 "hexagons" of diodes, layered on top of each other with tape and plastic between the two layers for insulation. If you look closely at the original schematics and reshape the diodes into a hexagon, then stack the hexagons on top of each other, you will see that the corners on the top and bottom layer of diodes line up in such a way that the wire pairs don't need to be split very much. One wires goes on the bottom layer, the other on the top layer with minimal separation of the pairs. So it's not as horrible as it looks, but it's limited how close I can get it with a flimsy circuit like this. hehe.
Yes I tried running a test on both at once, and also with only one computer connected. Doesn't change much whether I only have 2 or 3 devices connected.
But I don't think I will use this for the IoT stuff. I'll find another way to free up some ports on the router. I have 2 routers there. One for a point to point link which is always on (5 ethernet ports, but no wifi), and another tiny wifi router with just a single port that is only powered on while using the cabin. I can just replace this second wifi router with a bigger one with extra ports, and use those ports for devices like the TV and STB. Then have the IoT stuff connected directly to the main router that is always on.