They are everywhere, they're cheap and seem to work reasonably well. I can only assume that there are little to no surprises in terms of load cell selection. On a nice weekend I picked one up used locally and had a peek inside:

The load cells are rather basic, 3-wire elements and there are four of them wired together. The front ones also have push buttons stacked onto them to wake up the controller from power down.

This lack-luster PCB contains a COB microcontroller with a small 24C02 EEPROM an an LM324 quad op amp. On the left there's some kind of voltage reference and a bit of discrete circuitry I'll have to look at later, but together with R90 it'll be dedicated to generating the excitation voltage and temperature compensation for the load cells.

Usually load cells come as 4-wire wheatstone bridges like this one:

but the ones seen here are wired differently. They behave like potentiometers turned to 50% - give or take a few parts per million.

Measuring the voltages reveals how they are connected:

R   2.720V <->  R   2.720V

W   3.598V   W   1.841V

B   2.720V   B   0.961V

|                    |

B   2.720V   B   0.961V

W   1.840V   W   0.082V

R   0.961V  <-> R   0.960V

So the bridge is being excited with 3.6V from the reference and same sign outputs are connected together. The bridge output is at 1.84V and the 82mV on the low end feeding into 47R (R90) suggest the excitation is at 1.7mA (0.85mA per cell) which seems legit.

The cells will again be connected in this fashion when hooked up to an HX711 front end breakout. We'll have to check out what the manufacturer did on the temperature compensation side of things - maybe it's really just R90.