I didn't know lamp can be underdriven in A2
Are you talking about incand $urefire A2 ???
Now that I think of it, for some weird reason, the LED's themselves actually have 2 levels of brightness whenever one's cells are depleted enough that the incan lamp won't fire up. When one turns on the LED's at this point, if you fully press the tailcap switch, or full tightnen the tailcap, the LED's will gain a little brightness. I don't know why it does this, but both my A2's do this, and I've seen other comment their does it as well.
When you activate the switch to the first position, there is a 10 ohm resistor from the battery stack negative to the LED ring negative, and a direct connection to the battery stack positive. Close the switch all the way, and the 10 ohm resistor is replaced by a short circuit, or in other words, a direct drive to the LED with no extra resistance. You may be able to prove this to yourself, by removing the lamp from the head, reinstalling the head without the lamp, and activating the switch. In some A2's, the LED ring contacts will mate with the middle ring and body of the light even without the lamp in place, and you can see that taking the 10 ohm resistor out of the circuit by depressing the switch all the way makes the LED's brighter than they are with just the first position activation.
Any A2, flat or round has the same electronic/softstart regulator delivering constant 4V. Using 3.7V cells it won’t mean anything since bulb will still be regulated at 4V. LED’s on the other side will see more than triple current thus causing possible burn out (although i still don’t know for any case).
Yes, I've fairly sure I've seen a post discussing damaged (dimmed) LEDs in this situation.
I don't have much time for searching right now, best I could find at the moment was this (secondhand) report ...