So you can either implement optimal privacy hygiene, OR you can get access to services.
An awkward counterpoint.
This is actually interesting.
You basically executed perfectly on the supposed ultimate solution to staying in control of your data online, at the ideal level most wouldn't have the motivation/discipline to sustain... and it got your account revoked.
Of course it's extremely easy to understand why; you also happened to do everything a fraudulent user would do in trying to cover their tracks. And eBay's detection logic somehow figured that out, and I have to concede the point that it worked as intended. :(
This reveals where we're currently at: eBay can't tell you apart from a scammer, in the same way YouTube doesn't really know who I am and what I want to watch (the homepage feels like a hodgepodge of "throw everything at the wall" and more than anything else carries a fair bit of cognitive dissonance because it's kinda getting it right but just... not).
AI can't discern intention yet... because we're not feeding it enough data.
I'm not sure whether to be cynical, melancholy, viscerally depressed, angry ("man shakes fist at cloud") or what about the fact that the only way to prove individuals' legitimacy and authenticity is to basically mechanize and industrial level of private information sharing.
Naturally you've passed Facebook's bar for authenticity - it's analyzed the data you've provided explicitly, data you provide by interacting with the service, and data provided by those you interact with, and approved it - and so your account is not blocked there, and you can transact.
Considering the child comment you made about anti-fraud-as-a-service, it's kind of sad there isn't a way to allow eBay to access and evaluate the trust ranking Facebook has built. Kind of a "sign in with Facebook" sort of thing, except with a lot more data exchange going on behind the scenes.
A system like this would necessarily need to share a reasonable quantity of data, to overcome the same root-of-trust problems that eg makes fingerprint readers built into laptops generally worthless because they all merely send back a binary (yes/no) "OK, unlock" over USB after processing the fingerprint input inside the controller. (Yep.) Just sharing someone's email address and a binary "this person is legit" flag would expose eg Facebook to 100% of liability if that (entirely automated) decision was incorrect. This idea has almost certainly been thought of a lot and unilaterally shot down within 10 seconds. (Haha, it might even make a good interview question, whoever thinks it sounds good for more than 5 minutes maybe isn't great for a security position initially.)
Further, different services (Facebook, Amazon (online; Whole Foods), Apple, Netflix, Google, eBay, Walmart, etc, ad infinitum) engage with users in different ways, using different risk profiles, and the consolidated summaries of the risk assessments each company builds wouldn't be compatible with each other - at the policy level, because by definition this information is the proprietary, ostensibly-infinitely-valuable thing a company never, ever shares.
Getting something like this to work would be nigh impossible, and would hard-require both incentivization and regulation. I'm not sure how it would get off the ground even then, in much the same way the notion of a car would never pass approval today. (Hmm, but then we have Teslas driving themselves around off the back of utterly sketchy^W^WProprietary™ image processing, so maybe there's a twisted, messed-up chance, hahaha.)
I guess "interoperability" (in name only, lol) would basically look like companies sharing raw metrics and gross inputs and whatnot... oh man, it'd be a dumpster fire of incompatible implementations... and companies would absolutely try to get a leg up on each other by modeling what they shared in increasingly inscrutable ways that skate the line of the regulations.
Because at the end of the day internalizing anti-fraud-as-a-service is kind of a practical competitive thing to do.
Heh, the one thing would would definitely come out of something like this would be an entire ecosystem of startups promising to normalize the evaluation process. I wonder how long it would take for AI to bring us back to the broken position we're in today?