> The plane, after I eject, should do something reasonable.
There’s a good chance that it can't, and its not impossible that trying to do something reasonable combined with damage that led to and/or resulted from ejection could make things worse.
> starts broadcasting a mayday?
Great idea for peacetime over the homeland, maybe a very bad idea for military operations over contested or enemy territory.
> crashes into the nearest large body of water?
> attempts to fly itself back to base (we have the technology)?
If either of these are useful in a nontrivial share of ejections (except perhaps the former in conditions where it takes no special effort), then there is a serious problem with the training of the people pulling ejection handles and that needs to be fixed, rendering the action not valuable.
> Why would it be controversial to say "Look, guys, we should decide what the plane does after the pilot ejects. Maybe the best policy is just flying same course and speed until fuel exhaustion, but we should choose this policy, not default into it without consideration."
Because ejection is an action chosen when you can no longer meaningfully say what the plane does in any significant way. That’s the whole purpose. If it it is useful to address this question then you have a bigger problem that you need to urgently fix first.
Ejecting for for no reason would end the pilot's flying career. Ejecting for any reason will result in an investigation, at minimum. Pilots are expected to fly the airplane until the last extremity.
So while yes it's possible, it's unlikely, and the return on investment of making the plane able to do something like "return to base" in that circumstance would be a large negative number.
Even ejecting with good reason is enough to end a fighter pilot’s career. The rates for significant back injury are between 1 in 3 and 1 in 2 depending on the design.
Do you realise we're dealing with humans? Humans who make decisions based on a multitude of factors, or sometimes none at all?
Yes, but flying aircraft with ejection seats is demanding work and few humans are capable or qualified to do it. Most people can barely manage to drive cars safely.
I dunno, the current approach seems quite reasonable.  In the grand scheme of things the overwhelming majority of the Earth's surface is empty space where a plane crash is unlikely to cause much damage.  
You also have the complication that military pilots usually try to make sure their plane will crash in a "safe" area before they eject - many have died because they waited too long to eject trying to avoid a populated area.  Giving the plan a mind of its own after they pull the handle would be unlikely to go over very well.  I believe the scenario of a pilot ejecting from a perfectly good plane that keeps flying for more than a few seconds has only happened perhaps a dozen times in the entire history of aviation?  Not really worth worrying about.
I'll do my best. 
So you want to dedicate probably the rest of your career to automated diagnosis and recovery from crash conditions after ejection? Just so we can say we did a reasonable thing? Oh just the one case where the pilot rejects during level controlled flight you're saying we should be careful to let it continue on same course and speed? And if it's slightly changing course speed or altitude? Did we want to level out or continue the climb and turn? Do we attempt to maintain rate of climb even if it means throttling up? Descent?
The whole thing is so wildly ambiguous and niche that it's a black hole. When a pilot ejects the controller is gone. The controls are slack and it's just physics until fire.
If we had the technology to fly the aircraft back to base then there wouldn't be an ejection handle, there would be a "fly the aircraft back to base" button.
remember that part of optimization is the amount of resources spent developing a solution for a problem that just doesn't come up that often. in the microcosm of a single ejection there's probably a better way to handle it than to just let the plane continue on its course. in the macrocosm, there's probably better problems to deal with than the one that results from the relatively rare situation in the military and unheard of in the civilian sphere. it's also worth noting that ejector seats are explosive-assisted and any plane that's been ejected from is rendered structurally unreliable, and usually is so close to crashing that nothing can be done to save it even if saving it is viable. So most of what you do "in response" to an ejection isn't actually in response, it's about planning ahead. outside of a wartime situation where factors beyond your control tell you where you'll be flying, don't be in a place where it would be dangerous to bail if you think you might have to bail.