I’m not sure I completely understand the test, but if ESD protection diodes are involved, and if a peripheral has its own power supply while the I/O rail power is off, then the power provided by the peripheral can end up powering rails in the Jetson which are not expected. You might look at the power rails of “off” rails related to HDMI (such as EDID on the DDC wire) with a fast oscilloscope during the test and see if the ESD is adding power to a rail which you thought was “off”. You would possibly put the hardware in an invalid state (the “max” voltage would typically be the rail voltage plus the protection circuit voltage drop…if the rail is not powered, then the “max” voltage goes down by the amount of the rail “on” voltage…I/O would exceed voltages allowed because without rail power the allowed voltage goes down by a lot).
Note: I specifically mentioned the DDC wire because this is data provided by the i2c circuit of the monitor, but is normally powered by the HDMI cable from the GPU. Jetsons will only power this i2c circuit if it wants to query for EDID data. I don’t know the internal layout of ESD diodes in this, but if the rails are off, and if there is ESD protection, and if the ESD provides any sort of voltage to the DDC wire, then you would have to be careful with not exceeding a lower voltage. An oscilloscope on the voltage rail which is “off” might show a spike during ESD.