Your Site Reliability Engineering (SRE) team is managing an application deployed across three regions, utilizing Managed Instance Groups behind a global HTTP(S) Load Balancer. During the application of a critical security patch to Compute Engine instances, an error in the third region causes requests to fail. How can you minimize the impact on users from this unsuccessful patching?