INA230: About input sink current

Cafain,
The typical architecture of the current shunt monitor devices is to syphon some current off the inputs to bias the first stage of the amplifier, which is what allows the inputs to exceed supply. Usually this involves having a pair of resistors between in+ and in- to balance out the current coming from each input. See this short video for more information on the front end: training.ti.com/getting-started-current-sense-amplifiers-session-8-understanding-filter-related-errors
I expect that if you put 15V on one input and 0V on the other, you will basically have a resistive load from 15 to GND, so the current you can expect to see on the inputs would reflect this. Have you checked the IN+ pin current as well? 1.7mA from 15V implies around 9k differential resistance, which is in the ballpark of what most of our current sense amplifiers are. I can confirm this with our design team if you like.
On the data sheet, there is only a typical number of 10uA when Vsense=0V, and a plot later showing typical Ibias vs Vcm. We don't show the Ibias vs Vsense in this particular data sheet, but in the INA240, figure 32 shows the behavior of the input bias currents on that device as Vsense increases. Most of our products behave similar to this.