Sticking my 2p in here....
My money's on the hubs moving about in the dropouts though.
I'm not so sure - the spindle should rest against the drop out cut out - obviously vertical drop outs are better, then the axle nut usually has serrations that dig into the drop out, the whole caboodle kep in place by the QR. I need to measure it by the QR spindle but cannot be much more than 3 mm in diameter, that is not strong enough to hold the axle in place, and it doesn't - it provides the clamping force so the shear plan is not in the QR spondle but in the axle it self.
indeed, but in my case (vertical dropouts, IS mount and apparently crap skewers) the clamping force from the skewer just isnt enough to stop the left side of the hub axle from shifting downwards. its only fractions of a mm, and it takes a really hard, back wheel lifting kind of a stop to do it, but its enough to make the disc rub the pads. As the OP has found, leaning the bike over to one side counteracts this temporarily.
ive just looked at the dropouts on the forks and the paint damage and marks in the alloy back this up. the wheel has been moving on one side.
if you cant get to the bottom of it, i would lose some fluid too, means more lever travel til the pads bite, but will give more pad-disc clearance to allow for warps and dirt.