I think he means that this setup should work (I didn't really got to see it in your video)
. = channel
%% = pump
///////////////
. . . %% . . .
//////////////
This way, the magma never really touches the pump, since it drops down in the channel.
Please tell me if I'm saying something terribly wrong, just a noob trying to be helpfull.
I thought it was an unlikely idea, but amazingly, it worked. I don't know why, because the magma
doesn't only drop down in the channel.
Water on the upper level, sitting on top of 7/7 water on the lower level, would teleport via pressure to the first available place on a lower z-level. But magma doesn't have its own pressure, so the lower level fills only slightly faster than the upper level, and the tile behind the pump is constantly full of 7/7 magma.
I made a movie, but it is larger than DFMA's 1MB limit, because I watched the pump run for quite a while, expecting it to catch fire.
My original theory was that, since no other buildings are affected by magma in adjacent tiles, pumps probably weren't either, and their catching fire was probably a special case based on the fact that they were pumping magma, and not on the location of the magma. It's apparently wrong, because the pump ran for a while via manual dwarf pumping, and then I built a windmill and hooked it up to see if pumps only caught fire when running on power, and it still kept running.
I find this particularly baffling because the tile behind the pump had 7/7 magma on it the whole time anyway due to the no-pressure effect mentioned above.
Another experiment to follow.