0.3mm thermal vias in DPAK pads have failed badly

[ocset]

Hello, This thread speaks of using  0.3mm  thermal vias in the pads of DPAK power FETs…
We have 12 vias equally spaced under the drain pad of the DPAK

…we have now tried this, and unfortunately some (not all) of our PCBs come back from the PCB assembly factory with sharp “shards” of solder protruding out of the bottom  of the thermal vias.     
 
This is disastrous, as it pierces the thermal rubber pad that the PCB sits on, and contacts with the earthed heatsink. 
Some organisations  speak of putting solder resist on the bottom copper layer, with an opening just wider than the thermal via opening. I wonder if this will solve our problem? 
Alternatively, I wonder if we should just add solder resist over the entire bottom copper, so that the bottom of the thermal vias is covered up, and might stop solder from leaking out? 
I wonder if using leaded solder would also help  solve the problem? 
I also wonder whether the solder bath of the reflow oven was too hot?    Also, I deeply suspect that just maybe the DPAK FETs were actually hand soldered on, because of the difficulty in getting the reflow solder bath the right temperature for all the 0402 resistors, and also the big DPAK FETs. As such, the hand solderers (if they were hand soldered, we don’t know) might have been responsible for the protruding  solder?
-Maybe its just impossible to get the reflow solder bath the right temperature for a board which contains both ‘teeny’  tiny 0402 resistors, and also ‘big’  DPAK FETs? 
 

[dmills]

There is no solder bath in a reflow oven (That is wave soldering which is not normally used for smt)!

I generally tent the bottom, mostly to reduce solder thieving rather then for insulation (You could even try tenting both top and bottom, but carefully consider the mask expansion value given your vendors alignment tolerances if going there).

What does your stencil geometry look like?

If you are having solder puncture the insulator, I would say that your insulator is probably not man enough to be reliable over time and temperature and is most certainly something that would want extensive accelerated life testing. 

regards, Dan.