There are two good pictures to look at
and
The image of the cracked bracket shows the crack at a concave corner. That is an area of localized stress. It also appears there is a through bolt hole just below the crack. Presumably this bolt secures the bracket to the case. If this bolt is not properly torqued, it could allow movement localized to the concave corner.
Next look at the picture with the MTO "fix" (supporting bracket). The pre-rotator take off is relatively high, and it looks heavy. Think about impulse torque on the take off when the pre-rotator is engaged, but also think about engine vibrating when under power. Both will cause a lot of cyclical force to the top of the bracket. And if you look at the MTO "fix" it likely prevents horizontal or torsional loads from being applied.
Fatigue is the failure (by fracture) of structures that are subjected to repeated or cyclic loading. Look at the possibilities here: improper/insufficient torque of the bolt holding bracket which would allow more movement, a design with a concave corner (stress raiser), the skeletonized design of the part since removal of supporting material allows more movement and adds a stress raiser, cyclical load under normal conditions, and a impulse load (with torsion) on pre-rotator engagement.
Fixing any of the contributing factors will help. The MTO supporting bracket does this by limiting torsion and tangential movement. (It's also evidence someone at the factory has made an attempt to address a known problem.) Reducing the impulse from pre-rotator engagement will also help. The solid bracket will help in two ways: Less movement and removal of a stress raiser.
For me, it would be really interesting to see a video of the bracket while at cruise speed and when engaging the pre-rotator. I bet it bounces all over the place.
As far as a design critique. We have a cantilevered design with stress raisers close to a support point. We also have high cyclic loading. To me it should be obvious to any engineer that this design should be re-evaluated if failures are reported in the field. The saving grace is that this is not likely not a safety of flight issue, provided it is checked regularly.
I would suggest moving away from a cantilevered design, to a fully supported design, or changing to a material which handles cyclic fatigue better.