copy.fail patches can be applied with minimum downtime, and a VM reboots in 30 seconds, tops, regardless of size. I believe all the apex servers are configured as HA to keep the load distributed, so normal users won't feel anything when copy.fail is patched.

Our users didn't feel a thing when we rolled out the patches.

But the Ubuntu update servers are necessary to serve the update. Taking them down prevents the users from downloading the update. I don't know whether the update servers were affected though.