As an example for a typical explanation about red-eye reduction pre-flashes, the description in the Metz SCA3202-M7 flash adapter manual states
Red eyes are always the result of a physical effect. This arises whenever a person looks more or less straight into the camera, the ambient light is relatively dark and the flash unit is mounted on or directly next to the camera. The flash unit illuminates the back of the eyes, revealing the blood filled retina through the pupil. This is recorded by the camera as a red spot in the eyes.
The red-eye reduction function brings about a significant improvement in this respect. When this facility is used the flash unit triggers, prior to shutter operation, a few weakly visible preflashes which are followed by the main flash. These preflashes induce the pupils to close down, thereby diminishing the red-eye effect."
There is no significant difference in iris size: if any, there is a small overall reduction throughout the (chronologically sorted) sequence with photographs taken with a distance of about 1sec.
Nevertheless the last image not even looking in the direction of the camera (with the reflexes from the flash being almost or entirely off the pupil) clearly has the strongest red-eye, followed by the first image.
The middle image, in spite of having the flash reflex basically straight in the pupil, clearly has a minimal amount of red-eye.
And PostScriptum, to counter theories that some sort of reflection law ignoring the presence of the lens is responsible or that the area conversing with the blind spot is, here another photograph where I am focusing quite closer than flash and camera are but in a completely different direction:
What gives? The typical description of the red-eye effect strangely does not take into account that the eye is an optical system designed to create a sharp image of what you are looking at on the retina. If there is no overlap of the image of the flash on the retina and the image of the camera's entrance pupil (the aperture as viewed through the front lens), no retina area lit by the flash will be visible from the camera's entrance pupil and consequently the sensor.
So the main objective of a pre-flash to me appears to be making the eye bringing flash and camera in focus so that their respective images on the retina are sharp and disparate. While reducing the iris size would also help by increasing the eye's depth of focus (and thus decreasing blurring of the retina images), the effect seems minor compared to what the accommodation reflex can achieve in separation even with comparatively wide pupil.
Interestingly, in the fourth image the right eye on the left side of the image is considerably more red. Looking closely, the eye on the right side of the image has its vertical aperture narrowed by a drooping eye lid. This reduces the vertical blurring of the images of flash and camera on the retina, making them overlap less.
So contrary to conventional wisdom, staring focusedly at camera/flash (or a birdie waved at same distance or some other object that shares focus with camera/flash) will do most of the work even without preflash. An urgently blinking self-timer lamp should be almost equally effective for red-eye prevention as a pre-flash.
Now the people writing about cameras and designing red-eye reduction systems know their place inside out optics. It seems preposterous to assume that they of all people would not consider the implications of the eye being an imaging device after which cameras have been modeled complete with lens, aperture, and sensor surface. And money is riding on it: basically every camera review tests for red-eye.
What am I overlooking?