Not quite right.
With the original firmware the current through the switch LED is controlled only by the series resistor(s) on the switch PCB. A few k Ohms. The drive from the MCU is configured as a low-side driver (AKA “open collector”) which switches hard to ground.
The trick is instead for firmware to configure the MCU pin using the internal pulldown resistor, instead of switching it hard on. The internal pulldown is many times higher impedance than the external resistor(s), so provides a much lower current through the switch LEDs.
Serendipitously just enough to make them glow usefully.
This configuration can persist whilst the MCU sleeps. There would be other ways of dimming the switch using PWM but they would require the MCU to be awake, and taking much more power.
Tom E went to great lengths to minimise parasitic drain whilst the Q8 sleeps. It is more than good enough with the original firmware, though some of us prefer e.g. amber LEDs rather than green, and larger resistors/dimmer light.
But the possibility of using the MCU pin intelligently to give high or low levels to the switch LED is interesting, but yet another configuration option for a torch that frankly I can’t memorise. So many options, so little time, do you just want ramping (no) or modes (where do we start ?).
I’ve got the kit to flash it with e.g. Anduril, but I’ve never used it. Perhaps I should try, but these days I just like a torch with firefly/ultra-low. low, medium, high, turbo. Last mode memory, and a simple sequence to either start from ultra-low and move up, or straight into turbo and move down.
With of course a few hidden flashies, a short dazzling strobe burst to wake up dozy drivers whilst walking my dog home on dark country lanes. Not continuous, just a couple of seconds per button press. And a beacon flash that could last several nights if I got stuck on an expedition and had to activate my EPIRB.