For what it's worth, we're talking about mostly personal laptops, tablets, and smartphones here - not corporate servers. I wouldn't be so quick to dismiss the "physical access" requirement, although that does at least make it difficult for attackers to exploit the vulnerability at a large scale.
With the former, the device is already stolen. The only difference here is that the thief can re-sell it. That's not a huge win for the user who it was stolen from.
With the latter, you need ongoing physical access in order to install and maintain a presence on the device. That's a sophisticated attack, but sophisticated attackers already have ways of doing this.
> This makes it more attractive to do without stealing, but that's an attack style that is only really done by state level threats, and one that was still possible anyway, because hardware access is essentially game-over anyway at that point.
I don’t think that’s 100% true. Yes, you need persistent hardware to re-root the T2 on every reboot. But what about a one and done T2 attack - a drive by keylogger installer? Get credentials and then mount a more traditional attack to install APT/RAT.
Also from my reading of that, the T2 chip doesn't reboot with the laptop -- it stays powered on at all times after it first starts up. If that's really the case, a reboot won't do much except kill non-persistent things that are running on macOS itself.
So, a) a T2 hack persists until such reinstall and b) you don’t need a persistent beachhead if you’ve taken the hill and maintain persistence there. This exploit opens the door to that, and you can install keyboard intercepts via the new System Extensions Interface. Did it yesterday myself via an update to Karibiner. Notarization can be disabled by interrupting the network pathway and you can definitely disable SIP after you capture the password. In essence, T2 is the lynchpin that prevents the defeat of other aspects of macOS security.