I can remember playing Quake on a server with a latency of 40 ms and being an untouchable god because of it.
Even with compressed independent frames that never spend on future frames you get the transfer window lattency so you need a lot of unjammable bandwidth. Line of sight transmitters? Ir lasers? Kind of a hard problem though as you might possibly want to not light up the whole country with your communications.
And commodity hardware can't push frames without a lot of buffering. I guess they might afford custom display hardware but it's a complex problem too.
I guess it would be easier to just send the coordinates where you want everything dead and get the popcorns. Bonus extra is everything that actually is a threat but I doubt that it's a actually a big concern when doing CAS, although I guess a A10 pilot might be able to spot schoolyards and kids if they are not under pressure.
I read once that musicians need to be within about 20ms to collaborate in real time, and that sound travels through air at about 1 foot per ms. Fighter pilots can recognize an image flashed for 1/220 second (4.54ms) [1]. Etc. I'd call these operating parameters for drones, split between computing and communication.
1. http://www.100fps.com/how_many_frames_can_humans_see.htm Test #3
Realistically though, a thousand kilometers of distance incurs only about half a millisecond of delay. The real issue is probably getting the latency of everything else in the chain down to a low level. e.g. One problem commercial VR systems are currently engaging is getting latency down to a level that doesn't make people feel sick. The standards for a military jet are probably a lot higher.
On the bright side, a CAS fighter flying at a few hundred kilometers per hour is probably a much easier first-step for drone replacement than a supersonic fighter. I'm not saying the A-10 is obsolete yet, but it's likley going to be replaced by drones before any other manned fighters are.