You can not expect 20,000 lumen for more than 30 seconds or so from the MT09R. It gets massively hot. You would need a good liquid cooling setup to maybe run long durations.
I forget what the max continous output level is, maybe 3,000 to 4,000 lumen.
There are lights with fan cooling systems, but I’m not sure they can do 20,000 continous. 10,000 is possible, but that is a $700 light. The new Imalent MS18 might do 20,000 continous, but is it around $670.
Thanks for your advises guys ! 
Actually I use Olight SR90 but I need really more.
I dont expect 20K lumens for a long time but if y could get several minutes or even 10k for more coud be ok.
@Bluesword are 2 x Tracer batteries could produce enough juice?
I also could build a homme made multiple XHP70.2 trower as I already have the optical. But is this cheaper ?
Is this a fake https://fr.aliexpress.com/item/HaikeLite-MT09R-Super-lumineux-lampe-de-poche-3-x-CREE-XHP70-2-LED-max-24000-lumen/32855196030.html
or could be a aceptable upgrade ?
Thx !
The issue with the batteries is the voltage. You need a 2S voltage setup for the XHP70.2 MT09R. So a peak voltage or around 8.4V, the battery you linked to is a 3S setup with a peak voltage of 12.6v. It will kill the MT09R if you connect it.
Ailexpress has enough protections in place that there is very little risk in ordering things. You just might be out the money for a few months while waiting.
Yes I will definitely use a converter for dropping voltage.
I saw taht Imalent DX80 has very bad problem are these still present and is Imalent chronically a not reliable brand ?
Best !
I can’t comment on Imalent as I have never owned one of their lights.
If you will use a buck converter, you will need a really powerful one to handle the MT09R as it can pull upwards of 30-40A of current. Also if the converter has too much ripple voltage it can cause a feedback loop that can kill the driver worst case.
just have a look at the XHP70.2
to get 20000 Lumens with bypassed springs
about 6650 Lumen per LED we need about 10A per LED
3x 6,66Vx10A=200W electrical power to the LEDs so here about 170W heat
but looking on the FET just the DC power loss @4,7V gate voltage
https://eu.mouser.com/datasheet/2/427/sir800dp-93246.pdf
SIR800 —> 2.6mOhm —> 78mV —> 2,34W this will stress the MCPCB but should be OK for short bursts
this is why on the original topic i said the SIR808 is not optimal choice
https://eu.mouser.com/datasheet/2/196/Infineon-BSC009NE2LS5-DS-v02_00-EN-540021.pdf
Infineon FET —> 0,95mOhm —> 28.5mV —> 0.85W
I have here a HL MT09 XHP70 driver sent by accident somehow I can test the FETs gate resistance
I have measured the FET used on the red MT09R driver
its labeled like the original Vishay cant tell any difference
@4.7V Gate
I got 38.3mV 10A —>3.8mOhm resistance
@3V Gate
I got 165mV 10A —> 16.5mOhm
@2.5V which the original should have around 2.8mOhm as its gate voltage gets very deep till 2V pretty stable
I get only 2A with reasonable input volatge and the stunning drop on the FET of 1.1V
so here the FET has 0.55Ohm
Conclusion
the FET HL use is definately some sort of fake which wont work on any 1S driver well
with 4.7V on gate it has a usable resistance but still this FET is like the MCU fake
this is just the Turbo static power loss on 60% with switching losses this may be even more depending how fast the FET switches
the bad resistance on lower voltage definately points on a lot longer phase where the FET is not fully switched through
Lexel, thanks for confirming the FET is a fake one. We all suspected it was. 
Also, to clarify, only the red drivers from the haikelite factory have the fake FET’s.
The white drivers that I hand built are real FET’s and have not had a single failure that I am aware of out of the 100+ that are floating around.
I can certainly attest to this…
I have one of Texas Ace’s original mods and it is still my favorite and brightest light. I compared it to my new Emisar D18 yesterday and the TA MT09R blows it away, in both lumens and throw. It’s a much larger light, yes, but they are my only two lumen monsters so I can’t help but compare them.
I have 3 TA modded MTR09’s - they all remain completely trouble free and my favorite big light.
How much is TA paying you and how do i sign up? 
You know how it is, I liked the first one so much I had to have MORE! 
I spring bypassed my stock MT09R driver spring with 20AWG and also solder all around the spring (HK only applied a small amount of solder to 2 sides), I’m using VTC5A flat tops and my FET hasn’t fried yet, even did over a 30s torture test.
another member on here also used a stock MT09R driver to mod his MF01, he changed the stock leds with samsungs, bypassed the driver spring with 18AWG (even thicker than my bypass) and bypassed all his carrier springs with 20AWG, he is now getting 20,000lumens and has also done a over 30s torture test, his stock FET still survived
Wait, is there no coating on the driver spring?
That doesn’t seem too smart, since oxidation can mean contact resistance can shoot up, and heat up the spring significantly.
The stock springs had some form of (zinc?) coating. Just not the normal gold plating we are used to.
The stock springs should be using silver, as they say it in their spec sheet.
Zinc wouldn’t be too smart, as at high currents, galvanic corrosion would occur quite quickly, and would eat the phosphor bronze wire through.
IMO, it looks like there is no plating at all.
yeah, it might be silver, It was so long ago I forgot but I also thought it had no coating at first but when I asked about it they assured me it did but it was something we don’t normally see.
I know it was easy to solder to and showed no signs of corrosion so I saw no reason to question them about it further.
This sounds like great news for modders. 1.49W more available for the emitters and less heat generated. Do you know what the difference is between the BSC009NE2LS5ATMA1 and the BSC009NE2LS5IATMA1?