It is difficult to diagnose accurately without seeing an example of the failure and doing a cross section on it.
However, this problem could be caused by either circumferential or just large voids in the plated through holes of the bare board. This would allow the fibers of the glass in the laminate to absorb moisture, not just water but possibly acids or other process chemicals.
This moisture boils when the board is soldered causing the "explosion" like occurrence you describe. Some fabricators do a final bake to avoid this problem and in fact that is something that you can do to reduce the incidence of these failures.
Be aware however that baking might improve the apparent solder joint quality (the symptom) it will not fix the underlying issue of poor quality bare boards with voids or possible very thin or under spec hole wall plating, if, in fact that is the cause.
The fact that it is batch related makes me very suspicious of the underlying quality of the batch in question and I would certainly inspect the holes for voids and I would cross section a sampling of any suspicious areas for voids as well as for minimum hole wall plating thickness.