When I speak about this, the parameters of the engagements are narrower than what you think. This stuff is not for warring with the US or even Greece in 2030. Anyone with a decent stealth fighter squadron would clearly make this concept useless, or at the very least, greatly limit its efficiency. But stealth fighters do not exactly grow on trees do they? Most countries try to make do with 4th gen fighters and relatively older land based air defence. In short, pick enemies carefully. This is for crippling middling air forces.
AESA radar means seeing from further away, good missiles means shooting from further away, low observability means the enemy can't respond, and there is the slim possibility of dominating in dogfights with automation. A Kızılelma fleet will mean an "I win" button, if the enemy does not have proper countermeasures. And those countermeasures will be hard to attain. The bar for the enemy gets very high.
Technical issues you brought up are certainly problems, but not as much as you think they are. USAF already tested an AI controlled F-16 against a manned F-16, and unmanned one dominated.
The level of automation is the deciding factor here. Will the plane be able to;
- Spot,
- Identify,
- Lock on to,
- Shoot down
an enemy Su-35, while;
- Being at BVR distance,
- Without being seen,
- While resisting EW
If the answer is yes, this will be a monster.