The application running on these servers doesn't like the stun a snapshot creates, so they set aside a backup window during specific hours to allow a backup to run when the stun can be tolerated. Outside this window a stun will interrupt critical processes on the servers.
I see what you're saying with regards to backup windows, but that seems to be an awful complicated way to work around a simple problem. I could stand up dedicated proxy and repository servers just for this job, but that's an overly complicated solution as well.
A simple high-priority flag (for an individual VM in a job, or for the entire job) would make this much easier.