Scale operations within horizontal autoscaling enabled can be configured to add or remove only the appropriate number of containers to reach the desired running count without restarting all the other running containers in the service.
If you are using horizontal autoscaling on a background worker or job processing type service, you may want to configure it to use the restart-free variant in order to minimize interruptions of jobs in process.
The restart-free scale type may also be helpful if you are using horizontal autoscaling on a web application that can experience sudden traffic spikes, since the restart-free scaling will allow capacity to be added without interrupting all current connections. Additionally, web applications that experience strain on backend resources due to zero-downtime deployments can leverage the restart-free scaling so that horizontal autoscaling does not overwhelm those connected resources.

See the docs on Horizontal Autoscaling here and our updated guide for getting started with Horizontal Autoscaling.