The temperature coefficient of low value shunts increase, the lower you go in resistance.
Take a look at this datasheet:
http://www.vishaypg.com/docs/63216/VCS1625P.pdfIt is an excellent current shunt, I've used this myself, shipping thousand of product in spec. The values you are looking for are in Table 1, maximum TCR (typical is marketing bllsht). For the resistor above 2 Ohm, it is +/-5ppm, for below 0R03 it is +/-50ppm. And this is with the same technology and everything, only the laser/etching will make a different resistive element.
So with the same technology, while you can create a R resistor x good, R/10 is going to be much worse. And if you are able to handle the heat, by oversizing and cooling, you should. Also, load life stability is very very bad for low value shunts. For example there are Isabellenhutte shunts, which you can order with 0.1% initial accuracy, but just after 2000 hours under load, they can drift 0.2%. Which means you either calibrate very often or you go higher in the resistance.
At 100A, dissipating that 40W is not really an issue.