Just being the overconfident random internet person with ideas:
launch 100 vehicles, slow flying, potentially with gliding capabilities, converging from all directions onto a target at the same time, ditching main wings and piston engine at the last minute, firing a rocket to gain speed, and overwhelm air defence by sheer numbers. Combine low-flying attack profiles with "dive-bomb from above". Large sensors can be mitigated with mesh-networking between vehicles and sensor fusion, maybe different kinds of sensors on different vehicles. Slow-flying means they can carry heavier warheads thanks to more lift and less drag.
If slowly circling at high altitude, you also tie up resources keeping track of them.
Quantity is its own quality, and all that...
this is all without even bringing "AI" into the mix, but if you could, you can give them "goals" instead of targets.
It is possible in principle to build a loitering cruise missile with those features. It will not be small or cheap, at least not if you want something with the range and endurance to seek out and attack a ship on the open ocean in any weather conditions. Add up the cost and weight for all of those components you listed.
As a point of comparison, the latest Block IV Tomahawk missiles already do most of what you described. They cost about $2M each and weigh about 1.5 tons. Only the largest warships can potentially carry 100 such missiles.
Russia has used small, cheap cruise missiles like the Iranian Shahed-136 drones with some limited success against Ukraine. In a naval conflict such drones could have some value as harassment weapons against surface vessels operating in the littorals. But those drones are useless against moving ships over the horizon.
Those Shahed drones are too small. I'm just thinking from first principles, if an ultralight plane has a range of 500km and has a 100kg pilot in it which could be subbed for a warhead, you could do an awful lot of damage with hundreds of such things in the air.
Your numbers are way off. Ultralight airplanes don't have ranges anywhere near 500km, nor do they have the payload capacity to carry the necessary sensors and associated electrical generator. Ultralights are also barely faster than surface warships, and are too flimsy to operate in severe weather. Seriously, you guys need to quit watching silly scifi cartoons and do some actual math.
The Sadler Vampire was/is 100 kg something dry, 250 kg loaded and a range of 500km. How much power do you need for sensors anyway? Even with 100 kg for fuel, that's a lot of weight left for warhead and electronics. That's for a straight conversion of a COTS design. I'm sure corners can be cut for something which will only run for a few hours and never fly again.
launch 100 vehicles, slow flying, potentially with gliding capabilities, converging from all directions onto a target at the same time, ditching main wings and piston engine at the last minute, firing a rocket to gain speed, and overwhelm air defence by sheer numbers. Combine low-flying attack profiles with "dive-bomb from above". Large sensors can be mitigated with mesh-networking between vehicles and sensor fusion, maybe different kinds of sensors on different vehicles. Slow-flying means they can carry heavier warheads thanks to more lift and less drag.
If slowly circling at high altitude, you also tie up resources keeping track of them.
Quantity is its own quality, and all that...
this is all without even bringing "AI" into the mix, but if you could, you can give them "goals" instead of targets.