Without seeing all the code, I can imagine you can get into a situation where you are doing a 'software' reset while the interrupts and the pin level interrupt are still enabled, depending on what you are doing in the wakeFromSleep code.
Here is possibly something better-
void wakeFromSleep(){
//so we can get out of here (level irq fires as long pin is low)
detachInterrupt(digitalPinToInterrupt(pinSelector));
}
void enterSleepMode(){
// ... other code
pinMode(pinSelector, INPUT_PULLUP);
for(;;){
attachInterrupt(digitalPinToInterrupt(pinSelector), wakeFromSleep, LOW);
interrupts(); //I'm assuming irq's need to be enabled to wakeup, but I don't know
sleep_cpu();
noInterrupts(); //in case irq did not fire
delay(50); //debounce, change as needed
if( digitalRead(pinSelector) == HIGH ) continue; //if pin high again, do over
//pin is still low, so reset via watchdog - a real reset
wdt_enable(WDTO_15MS);
for(;;){}
//THE FOLLOWING code MUST be put in setup() to disable the watchdog
//after the reset (wd reset flag overrides, so must be cleared)
//MCUSR = 0 ;
//wdt_disable();
}
I don't have any 'older' avr's, and don't know much Arduino, but the code is probably close enough for an example.
Now you get a pin debounce- if the pin is not still low after the delay, ignore it and do again. You also get a 'real' reset where registers are in a known state.
edit-
I originally assumed the switch was external and used with the xlr connector, giving unwanted input to the switch pin when plugged in, but now it appears the switch is on the main board? (my bad) In any case, it doesn't hurt to get the pin wakeup working correctly.
I'm not sure it would make any difference, but you could also try setting those pinPotentiometerAx pins to OUTPUT LOW before the sleep. Depending on what is on the other end of those pins, 'clamping' these hard to ground may be harmless (causes no power flow) and may help in reducing any 'action' when plugging in. So instead of a floating input, where any voltage spike will have to find its way through the pin diode clamps and whatever else it can find its way into, you would have the pin output providing a direct path to ground of anything that comes along.
I'm sure there are more informed forum members here, so don't take what I say as being without any errors.