:)
I hope OP got in touch. Thanks for doing the thing.
Reading the thread made me think of a curious idea. It's got a couple layers of nuance, but could ultimately be really cool.
In short: have a proactively-updated list somewhere that shows the presentation and content of every single email that will be sent before someone is due to lose access to a given service or their whole account.
For example this could look like a (perhaps loosely-interconnected) web of support.google URLs with content like "you might receive an email like the following" exactly what the email looks like here> "and you can do" .
If this were a reliable resource that stuck around, I could totally see a community-maintained tool emerging that would monitor for these sorts of emails then eg send emergency SMS/notifications/etc so that carefully-coordinated action could be undertaken immediately by competent users. These notifications could be sent to a tech-savvy family member or associate/colleague, who likely would not otherwise have access to the account in question. This person could then gain impromptu access (perhaps via remote access) and take immediate action such as generating takeouts (and perhaps sharding them to raise the chances they complete in time), individully downloading or screenshotting critical info, and firing out bulk alerts reminding of alternate contact info. Basically disaster-handling to try to end-cap workflow disruption while the mess (hOpEfUlLy) gets fixed.
I can't help but think there's probably a disconcertingly large proportion of grandmas and effectively-illiterate users being caught in situations where they have very literally no idea how to proceed (eg, none of their Facebook friends are going to have a clue).
It would seem there is enough awareness and reactivity localized around HN to make a difference to that 99th percentile, if harnessed in a constructive and non-disruptive way.
It's a nuanced idea, but I'm reminded of the old YouTube story where optimizing the homepage made the site seem to load slower because it became viewable by a large demographic of people using sub-2G connections. Here, I can certainly see a temporary uptick in apparent legitimate account closures as the true false positive gap is surfaced, and that might translate to a bit of a suboptimal reputation hit. But in the ideal case, the new feedback would be integrated quickly, that transient spike would soon be forgotten, and in the long term, integrating the new data would help to further micro-optimize the performance of the 99th percentiles involved with the ML analytics processes in ways that exceed expectations.
I'm thinking of the bigger-bigger picture; I understand it's simply impossible for Google to provide tier-one support to every single one of their four billion+ free users. This approach could implement an unorthodox workaround that lets motivated individuals contribute toward closing the loop of false positives. Currently it is effectively impossible for Google's userbase to take charge of this particular edge case. I am sure the number of people who would be very happy to take preventative action against false positives completely outweighs (perhaps by an order of magnitude) those who posture and rage and make themselves look stupid in the forums when things don't go their way.
Knowing what all (or almost all) warning emails look like ahead of time is the only way a significant dent could be made in the false positive statistics; there's nothing stopping a detection project from getting off the ground today, but the lists of what content patterns to look for would have to be forged in blood by users who've had their accounts shut down, with this status quo compounded every time the wording of the warning emails changes.
The one caveat emptor I can think of is that there is of course no way to prevent malicious users from taking advantage of a community initiative that enables a heightened baseline of proactivity, especially if it goes viral and is easy to use. But I'm honestly curious... how might this be the end of the world? The worst the malicious users can do is download their data, which a) they probably have a copy of anyway, and if they don't, b) wouldn't be a great look in legal cases where a defense can say "here is evidence ABC used XYZ software to monitor their account for shutdown, which is why shortly after *point* this email was sent they *point* submitted a takeout request 5 minutes later."
I can definitely see some legal/regulatory/policy hesitancy (sigh), but (from my position as a relatively naive random internet user) I'm hopeful that hesitancy would mostly be without precedential substance, and this could actually viably happen with a bit of gruntwork and convincing.