CL in question: https://chrome-internal.googlesource.com/infra/puppet/+/41a3a0a6bfad8658f477fd89af343d2c45c35d63

This caused a 3 hour outage for the chromium CQ, and resulted in postmortem 97.

It would be good to understand why the folks on that CL didn't immediately recognize the problem once it was deployed and fix/revert it, and why no one on Labs was notified that windows machines were failing puppet runs due to this change.