Just throwing an idea. No time to test now.

If you can convince the IDE to add extra options to the compile command line, you could:

1. name __wrap_digitalWrite() your implementation of digitalWrite()
2. inside it, call __real_digitalWrite() when you want to get the implementation from Arduino core
3. add -Wl,--wrap=digitalWrite to the compile command of the final link step

Example wrapper:

void __wrap_digitalWrite(uint8_t pin, uint8_t value)
{
    __real_digitalWrite(pin, value);
    if (pin == 8) {
        Serial.print("Pin ");
        Serial.print(pin);
        Serial.print(" changed to ");
        Serial.println(value);
    }
}

Trying to put this inside a class won't help.

C.f. the man page of gnu ld.

Just throwing an idea. No time to test now.

Edit: I tested the solution below, and it works as expected, provided both the __wrap_ and __real_digitalWrite() are declared extern "C".

If you can convince the IDE to add extra options to the compile command line, you could:

1. name __wrap_digitalWrite() your implementation of digitalWrite()
2. inside it, call __real_digitalWrite() when you want to get the implementation from Arduino core
3. add -Wl,--wrap=digitalWrite to the compile command of the final link step

Example wrapper:

extern "C" {
    void __wrap_digitalWrite(uint8_t, uint8_t);
    void __real_digitalWrite(uint8_t, uint8_t);
}

// Calls to digitalWrite() are diverted to this.
void __wrap_digitalWrite(uint8_t pin, uint8_t value) 
{
    __real_digitalWrite(pin, value);  // call implementation from Arduino core
    if (pin == 8) {
        Serial.print(F("Pin 8 changed to "));
        Serial.println(value);
    }
}

Trying to put this inside a class won't help.

C.f. the man page of gnu ld.