There is no going away from this. Check it out yourself. These 10-second recordings from 1510, 1520, 1540, 1550, 1560, 1570 and 1600 should provide enough evidence. The first 5 seconds are recorded off the beverage, the last 5 off the QDFA.
So what brought these dramatic changes from last weekend? I don't know for sure. Maybe the radials I put out worked. Maybe propagation didn't favour the properties of the QDFA. Maybe there is some other obscure reason. My unqualified guess is that the radials worked. Btw I added another set of radials just before it got dark; they run perpendicular to the loop.
Of course this doesn't change the fact that the QDFA sensitivity rolls off from around 1000 kHz and downwards, and the fact that when the interference is reduced by approaching daylight the beverage is always more effective. But we knew that all along. The QDFA is set up for being able to DX the difficult midnight to dawn period. Thanks Dallas Lankford for this extraordinary experience!
Update: It appears that the effect of the QDFA bleeds away when civil twilight approaches. 45 minutes before sunrise (twilight lasts much longer at northern latitudes) there is virtually no difference between the QDFA and the beverage, although EU interference is still quite heavy.