AFAIK, the operator currently supports only rolling updates where Pods of a StatefulSet get updated sequentially, this basically adheres to the "no dowtime strategy":
https://docs.victoriametrics.com/victoriametrics/cluster-victoriametrics/#no-downtime-strategy

However, it might be hard and quite expensive to achieve in practice, e.g. #744, VictoriaMetrics/VictoriaMetrics#8591, etc. And the larger the cluster, the more painful it becomes, unfortunately.

Sometimes, it's more viable to keep VictoriaMetrics upgrades short and cheap at the cost of a small data loss. This is what the "minimum downtime strategy" allows to achieve:
https://docs.victoriametrics.com/victoriametrics/cluster-victoriametrics/#minimum-downtime-strategy

VictoriaMetrics is well-known for its flawless upgrade stability, meaning that upgrading from version A to version B of a given vmstorage node will almost certainly succeed without any compatibility issues regardless of the data inside. This is applicable even to downgrades, which is insane 🥇, thank you!

Thus, the idea is to make it possible to upgrade all vmstorages at once.