Spurious power button press with new Debian

Hello guys,
I have a custom carrier board powering my Beaglebone Black through the 5V supply pins on P9 (pins 5+6, I believe).

Because I’m having issues getting my working legacy Angstrom distribution image cloned onto new boards that have Micron flash instead of Kingston, I decided to give a newer Debian a quick try.
I have a 3.3V power regulator with an enable input on the board, so the BBB needs to assert a particular GPIO high to enable that power rail.
With the newest Debian distro, whenever I do that, I get a message that the power button has been pressed, and the BBB simply shuts down after that.

My 5V regulator can supply up to 5A, and this switchable 3.3V rail doesn’t even really have any load on it yet until I install the rest of the hardware, so I don’t believe it should have created any substantial drop in the 5V power.

I can do the same operation on the same (older) BBB with Angstrom installed and I do not have this problem.

Any ideas on what might have changed in the OS for this to happen?
I may go back down the road of figuring out how to clone my Angstrom build over to the new boards, but still kind of curious why this might have happened.

With the newest Debian distro, whenever I do that, I get a message that the power button has been pressed, and the BBB simply shuts down after that.

Any ideas on what might have changed in the OS for this to happen?

Code I could not say what exactly has changed, but I can say the code in the context of the button press NMI has definitely changed since.

I can give you something to experiment with that may actually fix that issue for you.

william@beaglebone:~$ sudo sysctl -a | grep nmi
kernel.nmi_watchdog = 0

If kernel.nmi_watchdog = comes back equal to 1 for you. Then simply changing this to zero ( 0 ) may solve the problem. so changing the cmdline= line in /boot/uEnv.txt to. . .

cmdline=kernel.nmi_watchdog = 0

Do keep in mind that this is supposed to be an x86 / x86-64 cmd line option only, so may not work. But the sysfs entries are there, so I’m thinking it should work . . .

Sorry, and actually there should be no white spaces in that period. So . . .

cmdline=kernel.nmi_watchdog=0