That was a good guess and nearly right. Over the decades I have learnt that the more attention I pay to power the better my project turns out. This was the case with valve amplifiers in the 1960s and it is still true today.
I have my Pi connected to a wired Ethernet and since it is a Pi 3 it only uses 4 of the 8 CAT5 wires for communication. Wishing only one cable to go into my outdoor case I fed 12VDC up the blue/whit and brown/white pairs into a buck regulator inside the case and from there I derive 5.5VDC for the Pi. I know the voltage is okay because the Pi is not complaining. So then I hit trouble and jamesh says
jamesh wrote: ↑Thu Apr 09, 2020 8:41 pm
Odd. The Pi should simply slow down if it hits 85, not crash.
Something is not right. I assume you are not overclocking?
Pi3 should be capable of running at well over 100 without failing (we make sure they don't though). I think we test to 120 or so in an oven.
Now I had never really suspected the Pi at all. I was certain that this was something that I had done either environmentally or with wiring so I looked
really closely at my wiring. The image tells the tale.
https://photos.app.goo.gl/EURFtQytAPYV5SXZ8 and proves my earlier point about power.
I have now removed the power from the RJ45s and use them exclusively for signal. I opened up the cable after the RJ45 and soldered a barrel connector to the blue/white and brown/ white pairs and plugged them back in. Inside the case the power wires are soldered to a PCB.
Both devices have come back up and have not failed throughout today but neither have hit CPU 85C either.
Moral: Do not feed power through RJ45s EVER. I should know better after the decades I have been doing electronics. I have now ordered some six pole (s'all I need) weatherproof connectors and will fit them when they arrive. As a temporary measure I have used DB9s.
Thanks to everybody for your help.
If anyone is interested, here's the entire project
https://photos.app.goo.gl/kCZFW5AQyCtS55vf7