I have done that. Comfychair rightly already said it will take direct drive from 2 18650 cells, I posted the mod a while ago: link. When bypassing the switch the led eats 8.5A from two very decent IMR's.
I’d need a much larger host to relieve the emitter of it’s heat if I were to consider 8.5A!
Thanks djozz, I’ve been studying your charts quite a bit lately! 
re: copper/brass/aluminum pill, the quality of the junction between the star and pill is likely more important than the materials. Brass might not be as 'good' as copper, but if you have a brass pill with the star soldered, and a copper pill with the star stuck on with Fujik or some other equally horrible crap, the brass pill will win. For an AL pill the best bet is probably a screw-mounted star with a proper thermal compound instead of any kind of adhesive or epoxy, no matter how fancy or expensive the glue claims to be. Copper pill with a copper star attached with lead-free solder and good thermal paste between the pill & light body would likely be at the top of the pile of course.
Not only does lead free solder have better thermal properties, its higher melting point means LEDs can be reflowed later without disturbing the pill-star solder joint. Plus, typical 63-37 solder paste has WAY too much flux in it for a large surface area joint like between a star & pill.
Anodizing does I think have a small amount of negative impact on heat radiation, but if the ano is black it more than makes up for it. Clear or light colored ano would probably be not very good and might show a small improvement if removed. But remember these are measurable but still tiny differences, ambient temps or wind vs. no wind will make much much more difference. 'Measurable' in this case doesn't mean 'big', just that you might be able to detect it with measurements.
This is all just guesswork based on how other stuff works, don't assume any of this couldn't be disproved with actual testing. :p
Even this horribly abused thing still works, electrically at least. There's just nothing to break like on the other LEDs - no skinny little bond wires, the dies are somehow a part of the base and not soldered like XML/XPG (I've had an XPG2 die lift off the base before, bond wires were still intact so I stuck it back in place with solder paste, and it worked again). None of the normal failure points exist with this one.
And yes, it's 2000+ lumens of very angry blue eye-burning light
Very cool 
I guess I noticed that the die was different, but didn’t put my finger on the missing brass wires. Very cool! The angry blue light, dangerously cool! Impressive findings and very helpful info on the pill/star junction.
I was planning on re-flowing the star to the pill and doing the emitter to star at the same time, applying some solder to the pads too while it’s all hot, would like to have the wires on there at that time but don’t see how I can accomplish that.
I do have one of TexasPyro’s re-flow masks to try out on the MT-G2, need some better solder paste though. I have some large diameter very old lead solder, should I grind some up and make my own paste? It’s probably acid core though.
This is a very interesting thread. Thanks to djoss' work, there are several direct drive MT-G2 lights documented in this thread:
So if I'm following this conversation correctly, I should be able to connect 2S Moli IMR26700 cells to a well heat sinked MT-G2 without damaging the emitter? Brand new, these cells can deliver 40+ amps continuous.
And my unprotected Panasonic 18650s can do +25 amps short circuit, yet only do a bit over 5 amps direct drive into a MTG2. The current delivered will depend on the voltage present at the LED under load.
OK Comfy. You convinced me. I will try it. If I don't blow my emitter (that sounds funny), I will thank you for your contributions to this thread. If I do blow my emitter, may the lumen gods forever curse you and your lights.
Use one for fins and the other for the cookies to eat while taking a break. 
I had overlooked that thread, nice collection of mods sofar , thanks for the link :-) .
As for using 2 x 26700 IMR cells, I'm sure because of the bigger size they sag less than 18650's, so they likely will deliver more than the 8.5 amps through your mt-g2, the question is how many more...
Oh, btw, at 8 amps you are at 85% of the max output (that is: if the heatsinking is done well), is there any point to go higher??? 
Wrong! Anodizing (clear or black) increases the emissivity of aluminum. Black is best.
http://www.infrared-thermography.com/material.htm
Also, brass is a horrible conductor of heat. Aluminum spanks its ass… it’s at least twice as good. And copper is twice as good as aluminum. I’ll take an aluminum pill/thermal compound any time over something soldered to brass.
Thanks TP, I had wondered at that since anodizing dramatically increases the sufface porosity and hence the surface area over polished aluminum.
Yes… shiny, smooth, and reflective are the enemies of efficient thermal radiation.
Then the opposite is true. Hence lapping.
As a matter of fact, every radiant heater I’ve ever seen has been shiny and smooth. Wonder why that is?
Surfaces in contact transfer heat by conduction, not radiation (BTW, heat radiates at a rate proportional to the FOURTH power of the temperature difference, and another fun fact… the noise of a turbulent airflow increases at the EIGHTH power of the velocity)
Radiant heaters have a shiny reflector because they are trying to reflect the heat from the heating element, not radiate absorbed heat.
Lapping enhances surface contact maximizing transmission. Radiation depends on maximizing surface area. Exactly why Alu radiates better than Cu I’ve no idea.
Where he bought quite a few oil filled radiant heaters, thinking that they would save on running the house heater. They gobble up electricity and radiate small amounts of heat in a confined area. Seemed pretty wasteful to me. They were indeed designed to radiate the heat of the hot oil filling them, and without fail they were smooth glossy painted surfaces. Attributable to poor design, no doubt. Inevitably the temp sensor would fail and they would sit and heat up to extremely hot temperatures, ready to burst and dump overheated oil in the house. They now sit in the storage shed but don’t let it get too cold this winter or his 85 year old self will drag em all back into the house!
This makes me wonder, we run tests and hold instruments close to a stationary light and figure we’ve got it all on paper. But when we use a light we are generally moving around handholding the body and of course most of the time it’s night out and the temps are cooler. So the ambient conditions when in use are almost always going to be more favorable to heat absorption than in testing, true? Will this make a big difference or mostly minimal difference? Obviously, going out on a 90º Texas summer night as compared to a sub zero gusty stormy winter night up North would have drastic differences in the lights operation. But just how effective are the fins on the heads of most of our lights, when the inner pill is seldom made as a press fit unit to maximize thermal conductivity. Most of the lights I’ve bought actually have dead air space between the inner finned areas and the aluminum pill conducting all the heat. Can we do more than wrapping the pill in aluminum to optimize the fins inherent to the outer head?
What if that contact was optimized, then a hole was bored through say 2 sides…off center of course so nothing internal would be interfered with…and an aluminum tube was ran through it and either left slightly protruding or shaped for aesthetics to fit cleanly into the lines of the light? Would this then allow an air path to conduct the heat straight away from the core? A heat pipe, if you will? Or would this too be inefficient without air directed through it?
The heat finds it copasetic and rewards us with happy escapes!
8)
<—-in need of pain meds, please forgive