This article is from 2017, but it seems the authors are not very familiar with how the industry already does it --- automatic probing machines can recover the netlist without needing to inspect the layers directly.
In fact, what they did, which is just image recognition, is probably one of the least time-consuming steps. Of course, it doesn't work for parts that have the markings removed either.
The paper, however, seems to be motivated for goals like penetration testing and MITM attacks on hardware _without_ destroying it-- like what "Agent Q" might do in a James Bond film (after Bond is able to pocket the circuit board and deliver it to the lab in an ESD safe pouch).
The far more common and boring scenario would be reverse engineering for the purpose of knock-off products. Not very sexy. Also, do-able in many cases by just noting a few key components, extracting eeprom/flash content and then consulting datasheets.
I'd recommend the whole podcast (Unnamed Reverse Engineering podcast) to anyone interested in hardware RE.