Hi, Gleb
You can't prioritise one endpoint over another - with Azure Load Balancer, as a workaround, you could have a Network Security Group rule assigned to the NIC of the other servers that deny traffic to the port you are after and then a Healthprobe alert that triggers automation to remove the block, but there will be downtime while the automation takes effect and the load balancer picks up the health probe change, or the alert starts the backend VMs instead.