[1] Well, there's a bit of ripple at about 248kHz, but that can be safely ignored. Both barakta's eyes and the flicker-o-meter are happy with the output.
I'm happy to accept either only slight ripple, or 248 kHz as perfectly good ways of preventing visible flicker. On the other hand, at 248 kHz, only a tiny capacitor needed to turn on-off control into slight ripple.
I've run LEDs with simple capacitor smoothing from half-wave rectified 50 Hz, and not been able to see flicker. Of course, I've overdone the capacitor so much that the lamp takes 5 seconds to fade out, so the reduction during 20 ms can't be that big.
I would guess that slight ripple could be seen by whatever the flicker-o-meter is, or it wouldn't detect full on-off PWM against any significant background lighting, so it could well be over-sensitive to ripple against real people's requirements.
Flicker-o-meter is a cheap handled oscilloscope kit (250kHz bandwidth or something rubbish) with a visible-spectrum phototransistor and some strategic black paint where the innards of the input BNC connector should be. Powered from a USB battery pack, it's portable enough you can point it at random light sources in the real world (some technique required to aim it off-axis with bright sources to avoid saturating the sensor).
It can detect slight ripple that barakta can't see, and of course any flicker in the 10s of kHz. But crucially, since you can see the waveform, you can reasonably infer how serious such things are. There have been occasions where she's had migraine symptoms without visual artefacts, and the meter has typically shown strong ripple from one particular lamp where others in a room are okay.
As I understand it, the way that flicker induces migraines etc is by turning off the light as the victim is trying to move their view to a new object, so the eyes have no feedback at that moment and overshoot. If the light is slightly dimmer at that moment, rather than off, the effect vanishes.
All bets are off with barakta. Bilateral
Duane Syndrome means that horizontal eye movements are achieved by turning her head, or by tilting her head diagonally and using vertical eye movements
[1]. Which means most of the time she has no saccadic suppression, leading to stroboscopic artefacts with any heavily-modulated light source. That's enough to give anyone a headache. In experiments she's been able to detect square-wave flicker at up to 6kHz (reliably at up to 2.5kHzish). There's more to it than that though, as her sensitivity has increased in recent years. (Devices with PWMed backlights or multiplexed LED displays that didn't used to bother her are now unbearable.)
The relevant medics are frustratingly unscientific about such things. All they know or care about is that "some lights give some people problems", and tend to recommend tungsten. (Barakta can see the 100Hz ripple on some tungsten lamps.)
I'm one of those people who can see flicker (at least PWM below 1kHz; I don't tend to notice ripple), but isn't particularly bothered by it, unless movement is causing stroboscopic artefacts.
[1] This seems to be her strategy for avoiding double vision while lipreading. Unfortunately, it means she's often bringing sources of glare into her visual field.