If there are indentations in the remaining solder, it sounds like the component has a solderability problem that is causing the component not to wet. This solderability issue could be made worse by the cooling fans dislodging the unsoldered component.

So start with the problem part and then work backwards to see why it's not wetting. It could be storage issues, contamination of the plated areas or lead / lead free solder mismatch.

Can you see any bulging or deformation along the bottom face (underside) of the components that popped off? If so that's a tell tale sign that you had a moisture problem.

If unsure you may want to experiment by baking a batch of the component PN in question prior to use. For the experiment be sure to use new components that were not previously exposed to reflow temperatures.

Bake them before use according to Table 4-1 of IPC/JEDEC J-STD-033B. If baking makes the problem go away, then you likely had a moisture problem and an audit of your MSD floor life tracking procedure is recommended. More information on this topic is available from Cogiscan's online Knowledge Center: http://www.cogiscan.com/

This is more likely a consequence of out gassing. The components, the PCB, or the paste may be moisture heavy and will need to be treated in an appropriate fashion to remove the moisture content.

Either dried in an oven [see Moisture Sensitivity articles for drying conditions.] or in the case of the paste conditioned and treated per the suppliers recommended methods.

In some cases such a phenomena will occur over blind via's and again is moisture or trapped gas in the via expanding explosively dislodging the parts over or close to the via.

Lastly I would check the cure on the solder resist, if it is not completely cured it too will continue to out-gas and create problems.