Maybe Eks has forgotten that he loaned me a Tek AM503 Hall effect current probe, which is exactly the right instrument to measure LED currents without introducing any series resistance. In order from the top, we have the amber, red 4-6, red 1-3, and red 7-9, with a scale of 20 mA/div (scope at 10 mV/div):
The currents increase by 10 to 20 mA during the pulse, which suggests that the 25 °C thermal change I estimated based on the forward voltage really happens. The power supply decreases pretty much as a 120 mV step and doesn’t vary all that much during the pulse, so I think the LEDs control the current.
Here’s the forward voltage drop screenshot again for comparison (without the amber LEDs):
Overall, they run close enough to 100 mA for my simple needs…