Delete instances in the managed instance group (MIG) one at a time and rely on auto-healing to provision an additional instance.

Perform a rolling-action replace with max-unavailable set to 0 and max-surge set to 1

Create a new managed instance group (MIG) based on a new template. Add the group to the backend service for the load balancer. When all instances in the new managed instance group are healthy, delete the old managed instance group

Perform a rolling-action start-update with max-unavailable set to 1 and max-surge set to 0