The WS2812 under the 5U4GB full-wave rectifier tube went into pinball panic mode:
It’s been running more-or-less continuously since late 2016, so call it
Because I’d be crazy to replace it with another likely-to-fail WS2812, I had to remove both of them before installing SK6812 RGBW LEDs and updating the Arduino Nano.
Unfortunately, I did a really good job of bonding the side light to the tube with epoxy:
The last tube manufacturing step involved flashing the getter onto the tube envelope, so as to remove the last vestige of air. Admitting air oxidizes the getter:
It was such a pretty tube, too …
The Smell of Molten Projects in the Morning 
Do you figure that stresses / uneven temperature from the LED epoxied to the glass is the cause?
Nope, this was entirely my fault. I tried to cut the 3D printed housing apart, but the stress from my diagonal cutters cracked the glass.
Drat & similar remarks!