Finally converged on an outdoor system where the camera & fan run off the same 12V input.

Haven't had a lot of luck with a general purpose splitter.  The idea with building a splitter into the camera connector was to reduce the number of connectors & hopefully make it more reliable.  Suspect the problem with general purpose splitters is some connectors tend to go unused & pick up dirt.

The fan converts it to 6V.  7V gave off a magic smoke odor.

Then the indoor system converts 5V .4A to 13V to power the big fan.  That way all the 5V drops in the apartment can power any camera.

Note all testing was done with stabilization on.  Disabling stabilization saves a negligible amount.  They're pretty useless without stabilization.  Stabilizing compressed footage in gyroflow isn't as good as stabilizing uncompressed footage in the camera.  Surprised the motion vector estimation didn't use more power.  It's all in the H.265 compression.

Another idea is applying the double sided tape to the chip-spacer interface instead of the heat sink-spacer interface.  Then apply paste to the heat sink-spacer interface.  When the heat sink bounces around, it wouldn't be smashing into the chip.

There's a wide variety of thermal pads.  The best ones have over $22 tags.

https://www.amazon.com/dp/B0BPYFXSVS

Lower quality ones have a $9 tag.

https://www.amazon.com/gp/product/B0CZRQR735

PTM7950 isn't made thick enough & is nowhere close to the best ones.  It's amazing how much the internet is manipulated to upsell the worthless.

Summary of thermal pads, 3 years ago.  Prices have doubled since then.  They recommend 

https://nabcooling.com/product/nab-cooling-thermal-pad-90x50/

It's not known if any of these would retain their thickness when the heat sink moves around.  It may not be possible to get the best dissipation without permanently fixing the heat sink.  It seems they all need to be oversized by .5mm to have any chance of a useful rebound thickness.

Another guy showed some thermal pads doing better than thermal paste.

Lions almost exclusively use 4k 4:3 24fps.  For that mode, it might be a bridge too far.  Higher frame rates & resolutions reduce stabilization & stabilization is the key.

-----------------------------------------------------------------------------------------------------------------------------------------------

Another outing in mid 90's killed it in 10 minutes.  So the failure point when outdoors is somewhere between 85 & 95.

Comparing 1920x1440 to 4000x3000, all this hardware definitely brings a big improvement in detail.

Slightly sharper results are possible by downscaling with nearest neighbor by a power of 2.

Made a video of stabilized 4k on different materials.