A scarfed joint is simply a slightly improved butt joint.
A scarfed joint is an improved lap joint !
The peak stress in the bond line is "slightly" improved by factors of 50 or so compared to a lap joint, and "slightly" improved by factors of 500 or so over a butt joint.
For those who understand metal:
If you do a riveted lap joint with one row of rivets it is bloody simple to calculate. (at least if you cover the excentric issue by corrective factors for your rivet strength, which is readily avaible in the MS standards or in the SRM of your aircraft). If you do a riveted lap joint with two row of rivets, it is still all easy, the load is distributed 50/50 to the two rows. However, using flush rivets the first row is still the more critical one, as the "bypass" load around the rivets lacks the additional material of the countersunk, but that is a relatively small effect.
Now it becomes tricky, if you do a 3 row rivet joint you no longer have an even load distribution, it will be an (around) 40/20/40 load distribution. If you increase the number of rivet rows, the loads of the critical rivets do not reduce further significantly. If you would do a 10 row rivet joint, the center 4 rows of rivets will be practically free of loads, while the first row of rivets will still transfer around 25% of the load.
This can be compared to a simple bonded lap joint, the edges of the bond line do carry most of the shear stress, the center area does not transfer any laods (and that is the portion perfectly protected from all environmental influences and hence the most durable portion...).
What would you do for a riveted metal joint? You would use stepped wallthickness, you would design "fingers" on the sheets to reduce cross section according to the desired load carried. This is exactly what you would do for a bonded joint, you adapt wallthickness to the amount of load you want each sheet to carry. If you adapt it fully from full thickness to zero (practically not possible), you get a constand bondline load, hence significantly reducing the peak stress at the most vulnerable edges. As a secondary effect you also reduce excentricity of the joint, reducing secondary bending stress.
Woodworkers do it that way for centuries, and it works.
If you look at an old, crashed wooden aircraft, you typically see the wood broken, not the glue joints.
Additionally you should always remember that there is nothing like a "continuous fibre", you always have some ruptures in the filaments so it is absolutely normal that indivindiual fibres do transfer their load to neibouring fibres via the resin. This creates the famous "inherent damage tolerance" of composites material. When single fibres fail locally, their load is taken over by other fibres. Already during manufacturing of the fibres you produce a lot of broken fibres, and during part manufacturing you add more of those. Additionally not all fibres are perfectly straight, perfectly parallel or perfectly tensioned, there will always be some fibres with "slack". However, load will always be transferred to the "best" fibres via the resin matrix.
In a bonded joint exactly the same happens. The only issue is to prevent the bondline from deterioration due to the environment and from secondary stresses due to misalignment, impact etc. Therefore periodic inspections of the repair may be required. Just like you would do for a riveted metal repair.