You're right on both counts.

My Mullvad comparison was focused on the credential, not on the LAN model. Brute forcing the account was never the issue, I missed the point about device-to-device was the real threat, fair point. I'm starting implementing LAN connections as opt-in, not default.

Second point, you're also right, because I was mixing prevent and identify, preventions shouldn't need logs, we can't take an IP address and a timestamp and attach a user to it, and the AUP imply something else, I'll rephrase this.