YES! I had this problem not one week ago. I ran a 3dmark loop to heat up the core in order to remove the GPU heatsink (same pink stuff you had) and then cleaned it very well.
I used AC3 (I use it on everything, i have a watercooled p4c 3.0ghz multiplier unlocked engineering sample running at 3.66ghz @ 45-48 under load, 39-41 idle) did a nice neat job even if I do say so myself and then put the ramsinks back on.
This is where thigns were starting to go south, I put the stock oem core heatsink on and put the plastic pins back in. The heatsink was moving, it was clearly not held down well enough. So I got nuts and bolts and made sure the core was very secure and all looked to be well.
Installed it, fired it up andafter about 10min the screen started to flicker (in 2d mode) so I went to check my temps and got this (see the screenie)
So I suspected i'd let a drop of AC3 bridge a couple of points and even though its not always the case it clearly can be conductive. But after looking for an hour I found no AC3 where there was supposed to be none.
Next step was to say f**k this, order the leadtek 5950 and some white paste and a pack of thermal pads to fix up the 5900 for use in another rig. Right now my 5900 is naked again and once more has been fully cleaned.
My advice is not to use AC3 but we'll see how the 5900 works when used with white paste and thermal pads over the next week (stuff due to arrive in the morning)
One pissed off AC3 user