To understnad scarf repairs, it is necessary to understand that the shear stress in a lap-joint is not unform. Way back in the 1930's Volkerson showed that the adhesive shear stresses peak at then ends of the joint and decay to zero in the middle of the joint if the overlap length is large enough. If the overlap length is too short, then the joint is susceptible to creep. In the early 70's Hart-Smith provided an elastic-plastic model which showed that for ductile adhesive systems, up to 80% of the bond strength was derived from plastic behaviour. Further, Hart-Smith provided design rules and equations for lap joints such that a designer could ensure that the potential strength of the adhesive was always greater than the design loads, or preferrably that the bond was stronger than the parent material, in which case the joint would never fail under any load case.
The theory behind scarf repairs is that if the scarf angle is sharp enough (about 1 in 30) the nasty "stress concentrations" represented by the shear stress peaks at the ends of a lap-joint could be virtually eliminated, giving a nearly uniform shear stress distribution. In practice, to avoid creep, the joint shear stress must be kept below the elastic limit. Given that such a large proportion of adhesive strength is derived from plastic behaviour, keeping the shear stress below the elastic limit for creep reasons means that the joint is not achieving full performance.
A further concern with scarf repairs is that the 1/30 scarf results in removal of substantial amounts of perfectly good structural material. If the structure was 0.5 inches thick, the repair would be 30 inches in diameter plus the size of the damage. You may want to grind out 30 inches of a structure but I certainly would not.
For thin composites, a lap repair can actually be stronger than the material. Hence the only reasons I would use a scarf repair would be 1. asthetics, 2. aerodynamic or clearance requirements on repair thickness or 3. stiffness restoration for stiffness critical structure.