Failover with GG based HA

Hello,

When GG is used as a HA or DR solution (by replicating changes to another DB that can be used in case of failure at primary), then how does failover take place?

I mean, I understand that target is open and available for the app to connect to and start processing, unlike Dataguard. But I am kind of getting drawn again & again to think of failover on lines of something like Dataguard. Lets say that the source (primary) fails, [server failure, network failure, instance failure, etc.] and the Apps connect to the target (standby). But all changes from source might not have got delivered (or applied) to the target. In that case what needs to be done? I mean what do other customers who use GG for HA do? Because those