“When you have an outage caused by a performance issue, you don’t want to lose precious time just to install the tools needed to diagnose it.”
Gregg wants sysstat, bcc, bpftrace and friends baked into every server image before the outage, not apt-installed during it. His made-up incident walks through a broken apt config, a firewall blocking outbound repos and an immutable filesystem, all while production burns. The tools cost a few megabytes. Teams skip them to save seconds of deploy time and then lose hours in the chatroom.