I just had the very unwelcome experience of a big RF module go sliding off the PCB after a couple weeks of light use. It is a Microchip RN2483 with castellated edges.
See all the edge pads? I would have thought it took a gorilla to pry off the chip, but (without dropping the PCB and) through normal use it came off after two weeks.
This is almost surely due to me not using enough solder paste before reflowing, and I use too little paste probably when looking at the small pads and thinking:
I better not put too much paste on there otherwise it will short to the next pad.
I guess half the solder moves up the leg, pin, or IC contact when reflowing. Only the other half stays on the pad and serves to hold the parts together.
My new idea is to apply more paste, even if it's overflowing and covering two pads. I assume that once reflowing begins, half the paste will disappear into corners rather than creating a bridged short.
To be clear, this would happen with any reflow and solder application procedure, like with stencils.
Anyone with more experience care to confirm or reject these theories?