I worked at RSADSI when I was a kid and supported the custom spin of TIPEM Hayden and Sophia used at Verisign. This brings back some very bad memories.
But... hopefully... people created overlapping windows of cert validity so there's always a valid cert available for their services and can tolerate the CA being out of action for 8(?) hours. Imagine if your TGS/Kerberos or AWS IAM IdP was down for 8 hours.
For persistent services using the affected ACME API, the window is usually 30 days.
But that didn’t stop Youtube and Youtube TV from going down hard. I imagine they’re provisioning ephemeral VMs or service instances and relying on them being able to get certs immediately, or something like that.