A video has 30 frames per second. A typical surveillance radar does half sweep in a second:
Basically, it looks towards the target, then looks somewhere else for 2 seconds, and then looks at the target again.
Radar uses wavelengths >10^5 times longer than even FLIR and, even with the best will in the world, you can't resolve finer than the diffraction limit (which scales linearly with wavelength). Sure, you've got a bigger detector, but 10^5 is a hell of a ratio to overcome.
You have the benefit of range data and doppler data from RADAR, of course, which is why you use the systems in tandem. However, that accuracy ain't that great.
I confess to getting curious about the depth-resolving power of the radar system whose spec is linked to above. That's generally taken to be: speed of light * pulse width / 2
? 3.*10^8 * 2857*10^-6 * (6/2500) / 2 1028.
That 1028 is in metres, as c is m/s, and pulse width was in s.
And the docs say: "Target resolution: 1000 yards". That passes the sniff test with flying colours. I get the feeling I can trust their other data. I now wonder whether the radar, at specification time, was designed to have that suspiciously round number as one of its requirements?