With one grain, you're right Xmas.
First, think impulse: m*v = I F*dt // Read that as integral of F wrt t
If v is the same at start and end (=0), then F must be higher some time to compensate for the time it was lower.
Before the sand starts to flow, the scales show the weight of the hourglass and every grain of sand. As the sand starts to flow, the measured weight will decrease linearily until the first grain hits the bottom. Just before the grain hits the bottom the measured weight will be original weight minus the sand in the air.
The measured weight after the sand has started to hit the bottom depends on what model you use for that collision.
If the grain is modelled to stop instantly, the scales will keep showing the lower weight except for some spikes when the individual grains hit the bottom. (The average would however be full weight,
including the weight of the falling sand.)
If it (more naturally) is assumed that that time it takes for a grain to "settle" after the collision is large compared to the time between two grains, the above measurent would be "low pass filtered". So you'd get the full weight (again, including the weight of the falling sand), but with a noise on top (I woudn't call it oscillation due to it's random nature).
When the last grain has started its decent, the scales will show an lineaily increasing weight. Increasing at the same rate as it was decreasing in the beginning. And when the last grain stops, the weight will jump back down to "full weght".
It's left as an exercise for the student to extend this to include the fact that as more sand gets to the bottom part, there will be a shorter way for he sand to fall, and thus less weight "in flight".
About the chicken: That's related to another thread here. You simply microwave it together with a reference glass of water. Then measure how much it heats up.
(I've already started filing all the monkey adoption papers.)