Hello Alan

You say that the solder slug is 5.3mm X 5.3mm and the finished drill size of the vias are 0.254mm.  I don't have a ready means to measure these dimensions on the Xray but that's a 20:1 ratio and in the Xray the solder slug/vias look closer to a 10:1 ratio.  Are the vias much larger that thought??

In any event, the particular part in the Xray looks good with only a few not too large voids. I'm also thinking that, in a failed part, the solder went down the vias.  That would  mean that the solder slug may not have been in contact with the solder on the pad; the solder slug on the component was not hot enough; or that the solder slug was old and oxidized.

As to the vias, I recommend the paper by Ning Cheng Lee at the 2012 Surface Mount Technology International titled "THE EFFECT OF THERMAL PAD PATTERNING ON QFN VOIDING."  It may be that the via size and distribution is contributing to the problem and this article discusses that.

There are 47 varieties of sparrow found in Africa, and there is insufficient data to calculate an average velocity for them.

Good Luck!