The consensus to take the block apart and have a look convinced me, so I did so. There was some black stuff in there, but not very much. Responsible for a few degrees, but surely not 50. One of the mounting screws was a bit loose though, so a bad mount seems likely. No decent guesses at how it worked loose yet. There's no impingement plate in my block as I'm not convinced they do more good than harm.
I discovered an even covering of the gallium, and no signs of the two blocks adhering to each other at all (somewhat to my disappointment). The waterblock was still shiny to my astonishment, I thought the copper would have tarnished by now. So the contact areas looked ok, but pressure was lacking.
System back together and reporting idle temperatures of around 50. It climbs steadily under load to the high 90s, and falls slowly when the load is removed. Temperature drops to 50 when the pump is turned off then climbs slowly, which I'm pretty sure means there's air inside the block which is displaced with the shock of the pump turning off.
I haven't done much about the loose connection on the pump yet, I think I'll take the wires out of the plug and solder them back in as a first attempt. I'm also (perhaps unwisely) using the unreliable pump in the cpu loop, as I'm more confident in the processor surviving a pump stalling than the graphics cards. What I have done is pull the pins out of the molex plug and shove them directly into the psu cable, which appears to have done the trick.
Good luck w3bbo, I'm not dissuaded from using multiple pumps in series either.
If it clarifies any of the above, I've split my system into two entirely independent loops. The one giving trouble contains a lapped ek supreme, three radiators, an xspc chipset block and an unreliable 18W ddc. There's also a horrifically crude if very compact loop with two graphics cards and a single radiator, which may prove to be so badly laid out that it's impossible to bleed it. I should probably have provided photos.
edit: A quick check, if the water level in a reservoir drops lots when the pump is turned on, this suggests an airlock somewhere through the loop is being compressed but not dislodged?
Loop is now definitely bled, yet the temperature still climbs steadily up to 100 even when idling. Radiators are not the limiting factor, so I'm starting to think pulling the entire thing apart is the way to go. Unfortunately.
