I always split jobs into multiple smaller ones!!!
For example: 100 server environment.
One big job, with one server failure is a 100% job failure
100 individual jobs with one failed job is a 99% success rate.
It's unrealistic to have to manage 100 jobs though.
Perhaps better grouping based on function and/or OS (e.g. SQL databases in one selection list, Exchange in another, etc, then the server OS itself in another)
From a very highlevel, applications like SQL, Exchange etc should NOT have AOFO enabled, but you do need it for the filesystems. Thus immediately, you should have two different jobs; one for apps/db the other for the OS/Systemstate.
If Win2k, you have to use VSP, since VSS is not available.
If VMware, it needs to be in it's own policy, same with Hyper-V.
That said, multiple smaller jobs is the best way to go in all cases. If not for configuration reasons (AOFO, GRT, VSS etc) then for operational SLA metrics, and even performance. As you can send multiple jobs to disk simultaneously.