Nobody has attempted to push what is possible with material science and aerodynamics because there is no point if the pilot wouldn't be able to handle the forces involved. With automation, those limitations no longer apply.
For an example of what I'm thinking of, look at air to air missiles. Sidewinder air-to-air missiles achieve an acceleration of 20g, far more than a human pilot can withstand. Drones in the air superiority role could likewise be designed more like larger, more advanced missiles than traditional aircraft.
I suspect that if engineers design combat aircraft to be pilotless, you'll soon see them executing maneuvers that would be far too punishing if the craft had to support a human pilot.
And remember too that the whole dogfighting concept, the getting behind the other guy for a tail shot, means less and less in a world of all-aspect missiles that can fire in almost any direction.
High-G maneuvers just are not a wildly important part of aircraft operations. Despite what you see in movies, high-G maneuvers aren’t what defeat radar-guided missiles and they’re not the bread and butter of dogfights either (which are all but a thing of the past).
They sure do look good at airshows though.
At the end of the day, if you’re a bomb or missile truck, you’re simply not going to outmaneuver a solitary missile that can perform 20G turns. You’re vastly better off minimizing your radar signature and taking out enemy radar sources using SEAD, at which point you can fly with impunity.