We aren't locked out of controlling machines in our network like Tor is. However, we are governed by a sophisticated design that is highly resistant to compromise without collusion, and rules that allow us to deal very politely with our traffic, while respecting the privacy of our users. First we separate the identity of the user from the actions of the user, using the VAULTS technology we developed. Secondly, we employ machines to notice "malicious traffic" and to take actions if they see any. All non-malicious traffic doesn't get logged. If the traffic is indeed malicious, it is a violation of the terms of service and is not covered by a secrecy guarantee, so logging can occur in that instance. So legitimate traffic can be attacked in two directions when we don't collude, from exit or from identity:
1a. People's Republic of Banana saw some exit traffic about civil rights that they don't like and want to know which account it originated from. Our machines didn't think it was malicious so it was never logged. The End.
1b. Or our machines DID think it was malicious but a human auditor didn't, so the log created was wiped. Identity of user never exposed to the auditor, regardless. Sorry, Banana, still don't have that info.
2a. Republic of Banana believes a certain person has an account with XeroBank, and wants to know what traffic is associated with it. First, that is against our policy, second, we are not subject to the laws or wishes of Republic of Banana. Third, we don't have a link from Deposit to Access unless the user created one. The end.
2b. Republic of Banana believes a certain person has an account, is performing "evil", gives us the originating IP address. If there is indeed legitimate reason to believe Republic of Banana, we can investigate if there is indeed "evil" traffic exiting our network that originates from that IP address. Then the situation goes to a human auditor. This again reverts to our privacy policy, and if the "evil" traffic isn't violating our terms of service or we don't believe it is "evil", the audited account will attempt to be informed (via email) that an auditor had reviewed their traffic at the request of Banana and found it to be deficient of any concern. Then it is up to the Republic of Banana to acquire court orders in all necessary jurisdictions, assuming legitimate user XYZ doesn't change to a new account. The message sent to Republic of Banana is don't waste our time with anything short of solid misconduct, otherwise you scare away the fish.