Largest issue would be false takeovers I think....have you looked at the 7.3.1 RLM-assisted failover enhancements?
(This sounds similar to a discussion we had recently with a customer who required sub 15 second failover times....challenging with the current active/active cluster failover setup.)