Our recommendation for production helm usage is to run helm template --output-dir manifests with all the values / cli args necessary for your environment and then kubectl apply the resulting manifests. This has the advantage of being able to review and/or version the manifests being sent to kubernetes as well.
Just took a look at the docs again - apparently there we suggest using the --dry-run option with helm upgrade which could also work.
Either of the options will avoid a restart of pods which helm would perform.