If I understand the paper correctly, this is a very poor technology for "normal" photography in its current state.

On the pixel level, there are no shades of grey. You get fully black or fully white pixels only. You have to calculate an average from a bunch of neighbouring pixels and blend together multiple "exposures" to get more data.

Because each pixel can only count at most one photon every millisecond, even the dimmest candlelit room at night would completely blind the sensor.

If you could have like a thousand times more pixels and a thousand times more samples per second, maybe it could be developed into a camera. I guess that would be way too expensive and wouldn't offer a significant benefit to most buyers.

The title doesn't make any sense - ignore it, it's just click bait. Which sadly worked as I read the whole article wondering the same thing.

Yeah. "Perfect" in the way it can tell the difference between two photons or three photons over a period of time, but almost useless for anything else.