I think transparent App failover is used in OPS where 2 or more DB with 1 or more CPUs are running off shared storage. If one server fails, the other one still runs and this config. won't cause any disruption.
connect time failover deals with searching for WORKING DB SERVICE at connect time. If one service is unavailable, then the other is tried, etc.
TAF (Transparent App failover) deals with rerouting active DB connection (i.e. called session ) to another service when original service fails.
The client application do not have to restart and reconnect, all client side DB connection statuses remains valid. Client even can continue fetching cursors opened before failure. Only current transaction is rollbacked.