Perhaps the elusive source of parasitic drain.
nothing you cant fix with epoxy
I do not know the temperature unfortunately, I am curious.
None the less it seems like the word around is throw that only thing it matters to LED life is how cool is everything (as it what is the host temp). While in CREE tests you can see how that is not real. Low current with huge case temp still outlives high current and low case temp.
I have seen it with my eyes how an XM-L1 started to shift at 45deg on an aluminum MCPCB, basically solder got into liquid state. Did not seem to have come with any issues of soldering, just by looking at it.
Solder is hard to know what it is from China, I suspect they use the 138C melting point solders (Bismuth based) in China just to claim lead-free and make everything simpler for them (less control over reflows especially comapred to other non-bismuth lead-free common solders)
However there are some claims only that due to certain reflow process such solder might happen that it melts at even as low as 120C
The other route of lead-free is the regular 200+C melting point solder which would probably create the need of too much attention/control
This is speculation of course, based on the fact I have seen an LED starting to flow around the MCPCB and it would be really hard to imagine 180-220C temps
I was thinking a bit about how hot an S2+ (unattended) gets in other driver situations, and for a rough approximation, once one value is measured like done in the OP, it is quite easy to do a basic prediction:
Thereās āNewtonās law of coolingā that states that the amount of heat transferred from one medium to another is a lineair function of the temperature difference. This law assumes a few things that are not completely correct in our situation: it assumes a constant heat transfer coefficient (unaffected by temperature), which is only valid for heat conductance. For heat convection, which also takes place in cooling of a flashlight, the law is only approximate, and for heat radiation (also present in our case) the law does not apply. Another assumption is that the temperature is constant over the entire flashlight which evidently is not true either. Still I can imagine that this law is not completely unusable for flashlight cooling, it might just be a pretty good approximation.
Some very basic calculations for the pill temperature following this reasoning:
*The OP situation: the amount of heat produced is the same as the amount of heat transferred to the air with a temperature difference of 90 degrees minus 20 degrees is 70 degrees Celsius.
*If 4x 7135 chips are used (so 1400mA), the power of the flashlight is about half (ok, that is simplified: the battery voltage is higher, so the power is a bit more than half, but then, the led is more efficient so more power leaves the light as light instead of heat, who knows which effect wins?), so in equilibrium, the temperature difference should be 35 degrees, so the flashlight will be 55 degrees Celsius (which is the human pain threshold temperature, the body being a bit cooler than the pill makes this flashlight always pleasant to hold)
*6x 7135 chips: in between, so 73 degrees Celsius. With a little added hand-cooling, or outside with lower ambient temperature, this flashlight may probably be used continually on highest setting.
*A direct drive FET-driver, say about 6A, which is double the power of the 8x7135 driver, so 160 degrees Celsius. Ai, that is well into the danger zone, for both battery stability and unsoldering trouble. But still, in freezing conditions with airflow, the light may stay cool enough.
Some day I will check one or more of the above situations in practice to see if the calculations make any real world sense at all.
Djozz, once the heat reaches the outside of the light you just need to calculate the overall heat dissipation from natural convection and emission. Just because for such a simple shape there are formulas and estimations for both. It is a simplified model but it shows how different surfaces react.
I cannot share the Excel file that I am using but if anybody is seriously interested I can explain how it is done.
Thanks for the test.
Iām surprised that there was a measurable temperature difference between coatings.
The above calculation is about as far as my simple knowledge of heat dissipation goes, but Iām happy to learn 
With the small difference measured, I guess that it can not for certain be attributed to anything specific, there will be small differences in how the lights are build as well (although I made them very close to each other). But as it so happened that the result (the yellow one a bit hotter than the grey one) feels logical I do not mind thinking that the difference is caused by the coating.
Very interesting thread!
This thread gets more interesting every time I come back to it
Hereās an old CPF post I wrote with an alternative approach to estimating what you can get away with in terms of runtime and temperature for highly-driven flashlights.
Thanks for all the work, reminds me of an episode of Mythbusters
Thanks for the link, I had never read that. Your approach is a bit different (how much time does it take before it is too hot to hold) which is very practical as well.
Nicely done. Nothing like doing a controlled test to help prove a theory.
Tbone are you taking into account the thickness of the surface? Calculations are hard to nail, but understanding the equations can be useful. Anyway, for conduction/convection, which matters most at non-crazy temps, thickness matters. It's not just about the "different surfaces".
Yes, I will. The thickness of the surface is one part of the overall internal thermal resistance. The result is a reduced surface temperature = less cooling.
I still have to get a feeling of how important each effect is. I need to crunch some numbers to see if everything fits.
Well the result of a thick surface layer is less cooling. But the result of a thin enough one less cooling, but not necessarily measurably less.
You don't a need full calculation to figure the coating effect though.
It looks like the thermal conductivity of some bad epoxy is around 0.2 W/(m*K), or resistivity is 5 m*K /W or 500cm*K/W. The surface area of an S2 is what 100cm^2 maybe (3*pi*10, rough guess). So dividing, the linear resistivity is abobut 5K/(cm W). Ok, for an 0.1cm thick coating that's 0.5K/W or 0.5C/W if you prefer. 1mm seems pretty thick, and 0.5C/W seems pretty un-catastrophic.
That thermal conductivity was pulled from some random article but polyester is only 4 times worse and air is only 10 times worse. That means having 1mm of closed cell foam would only produce a 5C/W increase. Ok, that's quite a bit for a 10W light, but that's 1mm of (non-convective) air! Real coatings are probably 0.1mm of something 10 times more conductive.
I think someone just blew up about nothing.
EDIT.. forgot the per watt in the last setp. that's important of course: fixed.
I know in myth busters there's always the one guy with the chalkboard and someone else is like, hey.. lets' just go blow stuff up and see what happens. djozz blew this one up real well already.
Djozz I actually believe in freezing temps it would stay cool enough with atmosphere and hand cooling. Its not apples to apples. I have a x6 triple xpl v6 3d DD. And around 60-70F+ within a few mins it gets scortching on turbo with 30q/hg2/GA etc with good cells. But when its 20-30F outside. I can hold it on turbo and never have to set it down. Especially if Iām walking around or just a windy. I know it has more aluminum but talking a lot more heat then one emitter could put out on dd
I just hope you understand that under cold ambient temperatures your DD is far from your DD when it is warmer or hotter.
Li-ion simply cannot perform in freezing temps, it performs very bad.
You wont get any high current at all that is why everything is so much cooler.
I just hope this graph makes sense to people, even after the cell warms up still heavily under-performs.
I get what your saying. I usually use 30q or hg2. But if Iām using for hours at the park I will put what I have with me in it. but use a combo of all three. My light isnāt at freezing when its turned on. It stays inside my house which in winter averages 74-75degrees. Its not kept in the car or anything. And if I carry its in my pocket against my warm leg. So its not the same as pulling a light out the glove box at 20 degrees. Iām not saying the light is running cool. Iām saying the light is tolerable to not have to put down. The head is still hot. But the handle doesnāt get to scortching levels at around 20-30 degrees. I have thick calloused hands from shoveling, weedeating, manual labor daily. And then from barbells in the gym. it takes a lot to burn my hands. When I put a light down itās hot.