We do get “failed” actions (We call deployment issues) but we also set retries 6 times with at least 6 hour between attempts.
This works for items like Office that will fail if the application is running. If for instance, a user is online and has Office open at 2 PM, the next attempt will be at 8 PM, when it is likely they user has logged off. If it still fails because the user simply locked the screen, many of our customers have reboots scheduled around 10 so when the action tries again at 2 AM, it should be successful.
I can’t really speak to SQL, we rarely have issues patching SQl and when we do, we pass it off to Server Admins or DBAs.