When does Leaf/Spine become practical?

I just started to research the next network upgrade and discovered the Leaf/Spine Architecture. It looks like a great solution but it might be total overkill for my environment.
I have less than 300 VMs on a collection of blade servers, along with the standard set of Firewalls, Load balancers, iSCSI SANs connected with 10 Gig networking. My VMs are static; no need to “spin up” a bunch to cover temporary increases in demand. I'm currently max'ed out on my 10Gig switches (about 50 ports) but I'm running multiple links to each server to separate iSCSI and data. If I make the move to 40/100 Gig I can reduce the number of cables/ports used by each server. I might be able to make the case for a complete rebuild that would include new servers and networking. <Tim Taylor grunting noises> So the question is… What/Where is the cost/benefit threshold for a minimal Leaf/Spine Architecture?