I'm not familiar with this particular camera, but I did look it up quickly and saw that this can run in a 3MP or a 1.3MP mode. Regardless, this is a 3MP 1/3" imager so it needs more light than, say, a 1.3MP imager. Smaller pixels = less photons per pixel in a given situation.
There are a couple of ways to deal with low light situations. One is to lengthen the shutter time. (I say "lengthen" because some manufacturers call going from, say, 1/10th to 1/30th of a second as "increasing" shutter speed which makes no sense to my frosty head)
The term "shutter speed" is a holdover from optical photography days. Nowadays there really isn't a mechanical shutter uncovering the sensor like we did back in the days of film. Instead, think of each one of those little tiny pixel sensors on your imagers as little bucket for catching light. Each bucket fills up with light during the shutter interval, then dumps it for processing (light level & color) so it can be converted to a number for that pixel on that frame of video.
In low light situations, there might not be enough light in all those little buckets to form a usable image. Again, one way to compensate is to lengthen the time the little buckets can fill up with light. It works great unless whatever you are imaging is moving, because the target's light will smear across many buckets creating what is being called here a ghost effect. (I would call it smearing instead) So lengthening shutter time can only get you so far. Some cameras have fixed shutter speeds, and some others let you choose an adaptive shutter speed -- often with a limit you can set to avoid smearing.
Another way to deal with low light is to use gain. Gain is an amplifier that effectively magnifies the signal coming off the imager. So what would be an unusable image is "boosted" into something usable. AGC stands for "automatic gain circuit" which means you let it add amplification at it sees fit. Gain is measured in decibels but, contrary to ADI's blog, you can't hear it or anything.
The problem with amplifying any signal (audio, video, wireless, anything) is that you also amplify noise. Video noise shows up as grainy little speckles that I call specularity but I'm open to that not being the correct term. The problem with video noise, besides looking crappy, is that all of those speckles are DATA. Darker image, more gain, more noise, higher gain = MORE DATA. This is why, contrary to what you might expect, your bandwidth goes up at night.
So which is better -- lengthening shutter time, or increasing gain? The answer is they are both good up to a point. Too much gain = specularity and higher bandwidth. Too long of a shutter time and you get smearing of moving targets. The trick is to use both as you can, and choose cameras that let you limit each technology.
My guess is this camera was set up for no gain and an automatic shutter speed that is being allowed to go too long which results in smearing. Then, when the illuminator is turned on, the scene has enough light for the shutter time to shorten and eliminate the smearing.
There are other factors that might be in play. Using a faster lens (that lets in more light) and making sure any iris is wide open would help as well. Switching to a lower resolution smaller image format would also help. Fewer pixels = more light per pixel. Finally, making sure this was a true night vision camera that moved the IR cut filter out of the way would be a good idea.