If something remotely viable and simple to implement could have been done, it would have been done by now. The Tomnod marketing excercise was rolled out in a couple of days.
Fact of the matter is that implementing and debugging algorithms takes some time and effort... and therefore manpower and money.
Add onto that you're expecting satellite companies to retask their satellites for a vast area of fresh imagery..... who's going to pay for that on top of the development manpower for your magic algorithm.
Your optimism and faith in existing "simple algorithms" is commendable, but that's all it is.
Tomnod was setup in the past for some Typhoon Haiyan and they were able to source/request fresh satellite passes from I believe the day it disappeared, seeing as they own the satellites. Their current effort is re-using an existing platform for a fresh task. Marketing maybe, but hardly a futile effort when at the time they were looking for a crash site, they had the satellites to do it and the technology platform to deliver it.
No doubt to yield good results it takes time to tune and improve the methods. All I am saying is that at a high level the effort to discount large sections of ocean appears trivial, and as I said I would expect this is probably something they are smart enough to have done already. If you are talking about recognizing an aircraft or specific aircraft from this imagery, yes that would be a significant effort.
Elastic infrastructure is easily available reasonably cheaply so I don't think some of what I suggested is that far fetched.