This afternoon one of our techs was checking load on kubernete's internal network. He was looking into a higher than usual load on kube's internal network now that we're 100% on 2.11 in us-east. When executing an eBPF program, it conflicted with a previously loaded diagnostics kernel module and segfaulted, whereupon the instance hung and needed to be rebooted. That was done and the sites came back. This tech was investigating on prod during the afternoon because we're unable to replicate the condition in dev (nor on Frankfurt which has been on 2.11 for a week). He's been given an informal verbal warning. Part of 2.11 is faster recovery. Previously we saw recovery times around 40 minutes and this time it was around 15. We'll keep pushing that lower and lower.