Talked with Lanning about the PR and what exactly using artifacts gets us here. Basically it sounds like the runners don't really understand what each other are doing (it's some kind of bank queue) other than "what step am I supposed to be on given what steps have already executed." As such you need to store the IP somewhere that's accessible to the other runners, as we're not guaranteed to pick up the other steps on the same runner PID. Artifacts solve that IPC problem even though that's not their primary purpose. There's probably a more clever way to do this, but good enough is, especially since the public knowing these ephemeral IPs behind devwall doesn't do an attacker much good even if they were to deduce the IP ranges we use internally from this data.

Uploading logs for failed tests as an artifact so they can be linked at the end of the run is probably a good idea as well, not sure if that should be a follow up PR or done prior to merge here.

Anyways, approved.

QA skipped.