While it is true that strong magnetic fields CAN deflect charged particles or magnetically vulnerable materials, these would only work against say ion beams or kinetic weapons using magnetizable metals for instance. Plastic or ceramic ammunition would pass right through, and since it is not the composition of the material that matters but rather the mass for kinetic kill weapons, these should still be effective.
For shielding I normally consider it to be a handwavium mass effect style potential shield, capable of robbing momentum from kinetic attacks and even laser light (maybe a combination of two or more different shielding technologies working together?).
As for armor that disintegrates under heat makes sense against laser weaponry. These would give off superheated plasma AWAY from the hull WITHOUT spreading it over a larger part of the armor.
Regarding EMP and microwave weaponry, while microwave weaponry is effective against organics, it is MORE effective against sensitive electronics. The electronics for ship control systems, target prediction and all that can be created simply and with resistance to interference far easier than the complex electronics needed for an AI. For example, your AC would be less susceptible to an EMP than an unshielded CPU. Both do their duty but one is just so much closer to the limits of what is physically possible that interferences can be that much more destructive. A single voltage pulse through an AC electronics would likely do nothing, while on a CPU can destroy it.
As far as I am aware, your microwave is not shielded from microwaves with a Faraday cage, the door for instance is shielded by the fact that the holes in the grid is smaller than the microwave wavelength(1mm-1m). That isn't to say that shielding isn't trivial. Either way, we already have a researchable technology that can reduce the vulnerability of our sensors to microwave weaponry, makes sense that this would also be effective to shield AI cores.
Aurora microwave weaponry I consider a handwavium for any anti-sensor and anti-tech weaponry, including microwaves and directional EMP lasers(some kind of ion magnetic beam? Monopole laser?).
Revisiting my on-off strategy for sensors, adding a delay in exchange for lower power use, I am reminded of two things... First, a large amount of power in most systems is spent starting up devices. My university for instance prefers to keep air-conditioners on rather than switching them on and off evenings and mornings due to the peak load being greatly increased when you start them up. For this reason a 10 second on, 10 second off strategy for power saving won't save 50% of the power, though a 1 hour on 1 hour off might come close, though possibly in that case, 10 min on, 1 hour off would be even better. Note though that a delay of even a second at these distances can be fatal in combat. Your targetting is already lagged by 1 second at 300 000km due to the speed of light and your laser will only hit it in another 1 second, which is a 2 second delay at those ranges. With an additional sensor delay of 1 second you potentially decrease your accuracy by 33%. Same for missiles that are not equipped with their own sensors. I would consider this power saving strategy viable for scouts where absolute real-time isn't needed, but it is quite simply not practical for combat.
I am still in favor of rock-paper-scissors gameplay. Every advantage has a counter. Even if we have to step away from reality just a tad to enhance gameplay.