The concept of death may be hard to explain because robots don't need to run 24\7 in order to keep functioning. Until instructed otherwise,a machine would think a person with a cardiac arrest is safe to boot later.
Actually no! Lower numbered laws have priority over higher numbers, meaning that if they come into conflict the higher number law can be broken. While the first law says they can't allow humans to come to harm, the zeroth law basically says that if it's for the good of the species, they absolutely can kill or otherwise hurt individual humans.
Robots or any part of an automated production line with a camera typically has a light as well to either see in low light conditions or to ensure it always sees with a similar amount of light hitting the lense.
Also, a lot of the machine vision systems I've run up against use red light, but it is kind of complex. If they want to detect say blood, I think blue light would actually give better contrast for detection.
They addressed this on the Orville. The glowing dots were not eyes. The droid had sensors that did all the work. The "eyes" were an aesthetic addition.
I really like the design of Assaultron from Fallout 4, they didn't have such issue because their eye is placed just above the glowy part, and the glowy part is the head laser that will one shot you.
To be fair it makes it harder to tell where the cameras are pointed (assuming they're not wide angle lenses and they're trying to work similarly to humans)
In the movie the bad ones didn't exactly disobey the laws, they merely found a loophole where by they can protect humans by taking over completely so "human error" couldn't harm humans.