I have similar anecdote, but with facebook: We had an web application get a huge bump in traffic as we approached the date of a massive live event that was being managed by the app. Suddenly, facebook shut off our access. Luckily, we had written the app in such a way that we could quickly disable the facebook features, and still have a functional app. If it had been critical to have facebook, the event would have been ruined, and our clients would have fired us. We contacted facebook, and our conversation went something like this:

FB auto message: our very amazing and awesome malware detector, which uses all this cool machine learning and is amazing, has detected that your site is malware and we are disabling it.

Us: Our site is not malware, it's just suddenly very popular. Please turn it back on.

1 week later

FB rep: You site appears to be on, please refer to the documentation for how to connect to our services, do not violate the terms of use agreement.

Us: Ok, it does appear to be back on now. Why was it shut off?

FB rep: You violated the terms of use agreement.

Us: How did we do that? Could you explain what term we violated, so we can modify our code appropriately?

FB rep: Do not violate the terms of use agreement.

Us: Ok, but which term? Are you sure your automated system didn't just shut the site down because it went viral?

FB rep: no response, ever again

We never changed the code, site continued to work fine. As a result, I live in fear that something like this could happen again, in a situation where facebook was a critical component. Did the ability to contact a human help in this situation? Unclear. They definitely did a good job of making us feel worthless and unimportant.