I didn't realize this design has been around for 10 years, loads of threads, comments and mods. I haven't really dug into it.
For whatever reason, if a shunt was overloaded and saw say +10V then that hits the op-amp input. Assuming the
MAX4239 has substrate diodes to GND and V
CC then I would say it dumps that to the rails, limited by R7, and the 3.3V rail which will just kick up and destroy things.
Other people can maybe offer their insight. Another USB ESD diode-array might be enough, after R7.
CSD16340Q3 N-ch mosfet is for 5V drive but it's getting 3.3V and I don't see anything special about it's leakage current spec.
I would not power analog circuitry from MCU power, it's just too dirty especially ESP32 does draw a lot of current on TX packets. To measure nA with a power piggy MCU right there I think is going to have some issues, EMI and more.