I got a bit distracted last time I tried to catch up on this threadā¦ but not today. Apologies for the delay in responding to posts.
Yes, the actual behavior is a bit difficult to put into a one-line formula for estimating power usage. When itās just one channel and itās running at 100%, it should be 350mA multiplied by the number of chips, like 5. But at lower levels or at mixed tints, it gets more complicated because of analog effects on the rise and fall of each pulse, and code which attempts to correct for that digitally.
It tries to keep the total power usage as close to flat as it can, when changing the tintā¦ but when changing the brightness, itās still pretty non-linear.
No, it keeps power pretty constant, so maximum brightness is approximately the same regardless of which tint is used.
The default is configured so the highest level can be used safely without overheating, since this item is intended to run for long periods of time without being touched and it has no thermal sensor to use for regulating itself. However, it also allows for a very simple hardware modification to run it even brighter, for those who want more light at the expense of having to be more careful.
Using fewer chips is also slightly more efficient at the same brightness levels, and allows for slightly lower low modes.
Keeping li-ion cells fully charged all the time is one of the fastest ways to reduce their life span. Itās not a good idea to keep li-ion-powered items plugged in all the time with the charge always at 100%, unless you donāt mind having to replace the batteries significantly more often.
It wonāt exactly kill cells overnight, but it can definitely reduce the life span from like 5-10 years down to just 1 year.
The pencil trick might not work. I never got it to work on a nanjg driverās stars, much less on a connection which actually needs to carry a reliable signal during use. But maybe I just need a different pencil.
Yes, it can be reflashed pretty easily. The code is open-source and the driver has programming pads and a standard SOIC8 control chip.
All modes share the same color temperature / tint setting. It can be ramped in almost any mode, by doing a āclick, click, holdā.
Candle mode has its own brightness level, configured by holding the button to ramp up/down.
That issue isnāt relevant for the lantern. It doesnāt have thermal regulation, and it doesnāt attempt to do any LVP while itās off.
Regardless, even if the lantern did those things, the issue was already fixed months ago. 
It should be possible in a D4. It even has a good number of LEDs for this purpose, so it can have two of each. Itād basically just need the MCPCB and driver redesigned to have two separate LED- outputs, and the firmware would need some extra code to handle four power channels instead of two. Not sure how hard it would be to fit everything onto a D4-sized driver though; itād probably be necessary to omit the RGB aux LEDs to make room for the extra power circuitry.
It is probably not over 1000 lumens. By default, probably about 500 or 600 perhaps. Itās hard to measure, due to the unusual beam shape and the black top surface.
The tint, however, is very nice on the middle color temperatures. The warm and cool LEDs mix in a manner which results in a pinkish tint instead of a green tint. This happens because the colorspace is curved, but the mixed tint line is not. The available tints span the pink line here, approximately:
