2) I see no technological reason that the systems memory be so unstable. So, you are making a copy of the pilot that is for all intents that pilot and purposely making it short lived. To me this would be analogous to intentionally cloning someone with a terminal illness that, while useful to you, will kill them in short order. What this would in terms of its legal status would obviously depend on how a polity treated sapience and rights inherent therein.
Well, there is the issue of managing entropy within such a system which has the tendencies of a quantum system whereby the scope of possible variables is endless (Turing's prediction of greater systems).
There is then also the fact that any sufficiently complex system can never be software-perfect, especially if it is undergoing ongoing changes. It can become more adapt but it can never become truly stable and it is impossible to compute when a program will halt (Godel's incomputability and Boltzmann's management of entropy).
Then there is the inherent issue of the mind itself. Specifically stimuli...
If a mind "wakes up" and realizes it is merely a ghost in a machine, you have psychohygenic issues which could lead to a negative performance impact: I don't want half of my processing cycles being wasted on "Why am I?" and hence the system is inherently unstable as so the echo or mirror never lives long enough to ask these questions: It is riding the psychological momentum of the pilot.
My solution, truly is to consistently reset and re-inform the system via long-term recall.
The name of the man is beyond my grasp at this moment but I recall the story of a piano teacher in the mid 90's who had only five minutes of short term memory. When it ran out, it reset because of the peculiarity of the damage left by his brain tumor.
He could teach every hour of every day: even into the night, convincing himself he'd slept in or some other nonsense and utterly believing it and he held a record of three days and sixteen hours of nonstop play with his only pauses being to read the letter he left himself explaining what had happened in the very shortest of terms.
He was permanently motivated and endlessly so. His skills even improved in long term memory as a pianist but he could not recall anything before the day whence his troubles began with the tumor. To him, it is still that exact same day, not an hour passed.
I use this peculiarity to my advantage here and also to avoid any messy laws around the copying of one's neural structure.
1) How fast does the initial scan take? I had thought that a ST scan (which is what this is essentially) took several minutes. I could be wrong on the time frame, but if I am not wouldn't this system represent a acceleration of that process by a order of magnitude?
The initial startup scan during scramble likely takes a good 60 to 180 seconds but the repeat "flashes" or checks take seconds and happen seconds apart.
How does the State handle this?
It would not: by definition, we are not handling an abstraction which can be extracted as an ST backup. This is a neural simulation. The machine has no legal rights and is in a "sleep" state.
It is not aware that it is, only that it is doing.