Keep reading, you’ll get there.
For FET selection for use in single cell (or 1s4p & etc) lights the RDS (on) vs VGS graph, or another graph containing similar info, is a primary factor for obvious reasons. Low RDS (on) is good of course… Clearly as long as we have a lower resistance than the equivalent wire length there’s not much gain to be had. When comparing SiR800DP and PSMN0R9-30YLD the graphs for RDS (on) vs ID (so top of page 3 and Fig10 on page 8 respectively) are very telling. The rest can either be seen in the specs table or with a scope…
FWIW I think you may have keyed off of comfychair’s comment. Remember that comfy’s very skilled but also hit and miss on driver knowledge - and just back from an approximately 1-year break from the forum.
I don’t want to sound mean about this, but we are quite aware that cells are a major limiting factor! For many of us this sort of information is quite mundane (especially in a thread like this one). You must have noticed all the discussion of high drain cells and the results they provide in ‘XYZ’ DD light? That’s not monkeys and typewriters…
Springs are another major limiting factor, that’s why we bypass them for these lights. You’ll find that if not bypassed at high currents they simply fail. Switches also have high losses, so some builders use the best switches they can. Another solution is to use momentary/e-switch lights where that’s no longer an issue (Tom E’s preference).
In terms of your burnt XM-L… you’re doing it wrong. I’ve personally measured an XM-L at 9A in short bursts - repeatedly. That setup went into a P60 which is still duty-carried with a cell that can make 9A through the LED. It’s got instant turbo rampdown, but the turn-on is always at 9A. I probably should have dialed it back, but I didn’t and it still functions. (Djozz has done considerably worse things, but not to the old XM-L.) Your missing ingredient is DTP. FWIW IIRC XM-L has higher current handling, old XM-L2 has high current handling, and new XM-L2 has mediocre current handling. (3 bond wires, 2 bond wires, 2 weaker bond wires)
At the end of the day, having some limiting resistance for a DD light is sometimes a good idea. You’ll see folks do it here. DD does mean that “any old cell” is never suitable - a [hypothetical] high enough drain cell will kill any 3v LED.
EDIT: Oops, looks like DB beat me to it. If you killed the XM-L while it was on a Noctigon mounted in a light… I suspect that your meter simply wasn’t fast enough to show the actual current. (Not enough solder on the thermal pad can also be a problem of course.) I added some strike-through since you’ve already ruled out my prime suspect.
EDIT2: I’ve destroyed more parts by “foolishness” than anything else.