I would guess the board has been screen printed with solder, and then the red glue was applied with a robotic applicator, but the droplet volume is too large for the small components ( or the operator mixed up the settings for the small components so they got a little too much) then the PNP placed the parts and then they were oven cured to cure the glue, flipped over and the top placed before going through either vapour phase reflow or dual side IR reflow. The bigger parts likely hide the glue spots well, and have the right volume of glue applied, but the small ones have way too much. Only reason they work is because these passed the power on testing phase, while those where the glue covered the pads and isolated it from the solder failed and were weeded out by ATE.