Specifications used in advertising for thermal imaging need to be understood.
The main feature in promoting a product is to state the
"Detection" distance. You must be aware it means just that, detection, and NOTHING else.
The normal standard is to
"detect" a man sized target (1.5 m x .5 m) at a specified distance. All you will be seeing is a few pixels of a "hot" spot, NOT that you will be seeing it as a "man".
From there you have
"Recognition" distance, where you can recognize that it is a man (or whatever) but not who it is.
Next, comes
"Identification" distance, often not stated, this is considerably less than recognition.
"Identification" is regarded only as identifying that a human "target" is a "good guy" or a "bad guy" not
who it is.

Another important piece of information is the operating or refresh rate, e.g. 7, 9, 30, 50 Hz and so on. What this means to the user is how smooth and realistic a moving image appears. Anything much below 25 Hz (Standard video camera frame rate)  produces a lagging image with a moving target, especially a fast moving one.
At first thought it would be fair to say that a "new" picture, say, 9 times a second, a commonly available rate, would or should be fine.
Sorry, no it is not, OK maybe for casual observations but not much more.

The next thing to consider is the detector pitch and FPA (Focal Plane Array) size, thermal devices are commonly 25 microns with an FPA  of 384 X 288 pixels. While this is quite reasonable for most work, it can be a little coarse as the image suffers when using "zoom" magnification settings.
 An FPA at 640 X 480 at 17 um compared with one of 384 X 288 25 um, the magnification will nominally be the same but the difference in resolution is significant. Admittedly it may hard to justify the extra cost involved.

Now if you have an FPA of 640 X 480 with a pitch of 17 microns and one of 384 X 288, also 17 micron, using the same focal length lens, you will have approximately 50% higher optical magnification with the smaller FPA. This is due to the larger FPA producing a wider field of view which in turn produces a larger overall image. Viewing a specific "target" within that image, at the same distance with the same lens, either size FPA, when compared optically, that specific "target" one will contain no more or no less pixels than the other. Looking at the provided figures, the 640 X 480  FPA "suggests" it will contain more pixels  in the "target" image than the 384 X 288 FPA, but this is not so. 

The smaller FPA will have  slightly​ more degradation in resolution due to the greater expansion to the viewing OLED screen, but the extra magnification will more than offset any loss of resolution.

Comparing the 640 X 480, 17 micron and a 384 X 288, 25 micron FPA the magnification is practically the same, as the overall FPA physical size is not that much different to each other. So therefore the optical magnification is similar.

When it comes to visual "assessment" of a newer model thermal of 17 um 384 X 288, compared with an older one of 25 um 384 X 288, do not expect to see "high definition" images. The difference in resolution, while certainly improved, is only about 40% better. This amount of improvement is worthwhile and very noticeable but it will not be stunningly better when seen by eye. 

There is an odd manufacturer or two advertising 12 micron pitch. Be aware that they can have an FPA so small with such a low number of pixels, that when you expand it to a point where you have an image large enough use, it has no better overall resolution than a unit with 25 micron pitch!
You need to note the pitch of the detector and not confuse the pixels specified with that of the "viewing screen" specification.

There is a supplier advertising a thermal clip on unit with as few as 160 x 120 pixels and 25 micron pixel size, and this is coupled with a 9 Hz refresh rate. This would have to be the most unsuitable specification you could possibly use for a rifle scope. Apart from the very coarse resolution it would produce, the 9 Hz refresh rate could easily mean a complete miss, or worse still, a wounded animal, due to the lag in the image. At the very best 9 Hz is only barely suitable for observations as mentioned above.

Sensitivity and the lens. The "f" number of the lens governs the amount of "light" entering the unit as it does with a camera. Top end thermal units generally use f 1.0 to  f 1.3 lenses against higher numbers of lesser makes. System sensitivity is more important to the user than just FPA sensitivity.
System  sensitivity is determined by multiplying the detector sensitivity in milliKelvins by the the lens "f" number squared.

For example,  a unit with an FPA sensitivity of 50 mK using an f 1.0 lens will have an overall sensitivity of 50 mK. e.g, 50 x 1.0 x 1.0 = 60, while an f 1.3 lens would be 50 x 1.3 x 1.3 = 84 mK. If you jump to say, an f 1.5 lens you will now have a system sensitivity of 112 mK. This shows the decrease in the system or overall sensitivity with the "slower" lenses. (Formula supplied by people far more knowledgeable than me). 

Don't be too alarmed about overall system sensitivity loss, (within reason), with other than low specification products, while it is certainly reduced by the use of "slower" lenses, the feature that is reduced is the ability to see small differences in temperature, eg such as shading on an animals body. An animal for example, will still stand out against the general surrounding country. There will be some "visible" losses of lower temperature things like trees and shrubs as they cool later into a cold night and early morning but the detection distance for animals will not be of great concern. For example, if you have a detection distance of 2750 m for an  f 1.0 lens you would only drop to around 2200 m with an f 1.3 lens based on the "Johnson Criteria" figures. BUT having said that the highest sensitivity should still be be strived for where ever possible.
Kelvin, just to explain -
One degree Kelvin is equal to one degree Celsius, they just start at different points.
E.g. 0 degrees C = 273.15 K, and 1 degree C = 274.15 K, both up by one degree, and so on, so you can see from that, in each system  one (1) degree is the same. Kelvin is the scientific system for absolute temperature measurement. 1 mK is 1/1000 of a Kelvin.
So with an overall system sensitivity of
60 mK for example, you can detect temperature differences of 60/1000 or 0.06 of one degree of temperature!

Now having seen what provides sensitivity, resolution etc., ever considered why some thermal items are so much cheaper than others? The lens is germanium, NOT glass, and they are expensive to make. Some manufacturers use lenses that are very suspect, this is usually due to it being poorly ground to cut down on cost, and that causes the passing light rays to disperse slightly and not focus exactly on an infinitely small point at the focal length,  which can NEVER produce a clearly focused image regardless of how much you try!

 Do NOT believe all the claims made by most suppliers, that you can see through fog, smoke, rain, etc.
Well, OK, yes you can, but with reduced performance (Detection distance). Remember, anything, including foliage, that completely prevents heat being "seen" will prevent detection. Even humidity will alter the performance capability of thermal imaging (And all other types of NV as well).

The real benefit of thermal imaging over other forms of night vision is that you need NO light at all, visible or infrared, and at the same time it is possible to see objects that are hidden in shadows or partially obscured by foliage.

As just a point of interest, thermal will have "white hot" and "black hot" options, white hot is great for scanning for a target, but once an object of interest is located, changing to black hot tends to sharpen the image outline. High end commercial and professional makes will change the reticle "colour" according to the white hot / black hot polarity. e.g. black reticle on a white hot target and a white reticle on a black hot one.

On the subject of black / white hot etc., colour "palettes" are available with some thermal imaging products. While these may seem attractive to some users, it is my opinion that they are more of a gimmick than a real benefit. I have some units that have this option, but in general when I use these, they stay on the black and white setting, only changing the option of "black hot" or "white hot". Basically all manufacturers provide this option for what it is worth. 

Be aware that thermal imaging rifle sights cannot be exported from the United States of America without a US Government permit due to ITAR regulations. Those that carry an American brand name and are available in Australia for general purchase will be re-branded Chinese, or similar, manufactured items.