The false positives at scale problem isn't something that could be solved by an appeals process, because the problem is a step above that: a lack of continuous process improvement.
Why was the false positive generated? Why was the user unable to use the process in place to appeal? And then improve the root process cause.
This is what seems to be broken at Google and Facebook, and is the fundamental difference between a black hole support model (make problems invisible to us, because they never get to us) and one that works at scale. You're always going to have false positives, but I don't think anyone feels like the mega-corps actually have a process in place to monitor and improve the process for dealing with them.
Algorithmic moderation will always have false positives. The question is how you deal with them: Do you have an appeals process that involves human review or do you just say "sucks to be you" and let the false positive be the user's problem?
In a similar way, car accidents will always happen. However the car industry decided (not completely on their own) to invest in safety so that when accidents do happen, they are less deadly.
> Anybody at any internal discussion would probably disagree with this, instead saying something along the lines of "there may be deficiencies with our support process but we can always improve them later on."
The mythical time of "later on". When exactly would that be?
And yes, you can always improve anything. This is a vaguely pleasant-sounding non-statement.