"I would liken it to a car," he says. "A car on the road has many parts that facilitate its movement - the gas, the transmission, the engine - but there's only one spot where you turn the key and it all switches on and works together. So while consciousness is a complicated process created via many structures and networks - we may have found the key."
...objects that could be as much as 8 wavelengths of that monochromatic light
not 8 wavelengths, but "8 orders of magnitude bigger than the wavelength of light used in the imaging process" (from the linked article)
Bag: Should be avoided. Baby: Should be avoided.
Bag: May involve gently changing direction, do not brake erratically, do not disturb flow of traffic. Baby: May involve driving into the ditch, other traffic, making full use brakes, honking horn, etc.
And a bag with a baby in it?
You are missing the point - the system should not be trying to model the value of the thing on the road - at best it should model things for their expected behavior. Is it stationary or not? Is it likely to move or not?
If you have enough information to satisfactorily "report your lost Bitcoin", then you have the bitcoin itself.
For a bad analogy, think about a 20 dollar bill. You can't go into a bank and say "I a 20 dollar bill, can I have another 20?" and expect success. If you go into the bank and say "I lost a 20 dollar bill and I remember the serial number, can I have another 20?" you still won't succeed. If you could somehow go into the bank and say "I lost a 20 dollar bill and I remember the serial number, and I can prove that it is mine, and I can prove that I haven't already spent it, and that no-one else can spend it" then maybe you'd have a chance.
Once you can do all of that, you have enough information to "report your bitcoin lost"... but that means that you have enough information to rebuild your bitcoin.
Numeric stability is probably not all that important when you're guessing.