Well, really hot.
This diffuser tube from TCSS was in a 1" thin wall blade with no tip and no other insulation. The blade only has 55x2 pixels, and was operated for about 10 minutes at full RGB white.
The temperature reached about 100ºC / 212ºF, well known as the boiling point of water.
My first hint that things were getting spicy is that I touched the pixel strips and it hurt. Then I checked the temperature probe and saw the number. (My test blades all have embedded temperature probes)
After it cooled down I started another experiment and realized that the pixels were much more visible than they should have been. At first I thought the strip must be pressing against the side some how, then I realized, no, there was literally no diffuser in that area.
Once again, this was a 55x2, minimally insulated, open ended blade.
Part of these experiments is to demonstrate the characteristics of the blade under extended loads and such. More voltage and more current is not more better. The relatively extreme temperature causes the colors to shift due to quantum effects, and the brightness is not significantly greater. As Andras (aka protonerd, aka Obi1) has first showed years ago as we were developing the original pixel blades, the ideal voltage for pixels is around 3.7v.