[pdonis]

> my understanding is it's primarily because of how adversarial fraud detection is.

Let me rephrase this to make it clearer what the actual issue is: it's primarily because the cost to Google of false negatives on fraud detection (failing to detect actual frauds) is much higher than the cost to Google of false positives on fraud detection (flagging users as fraudulent that actually aren't). So Google is willing to accept a large number of false positives in order to avoid false negatives.

Or, to put it more simply: the incentive structure of what Google has chosen as its core business model means it is in Google's interest to randomly penalize a large number of bona fide users simply because a much smaller number of users are fraudulent.

In other words, this is basically unfixable unless Google changes its core business model.

See the problem?