From what I read there was a fracture thru the rim of an output stage planet gear that originated at a spall on the spherical roller bearing outer race surface. Normally, bearing race surface spalls are fairly benign events that propagate slowly, and do not result in a rapid catastrophic failure. They also produce ferrous debris which is quickly and easily picked up by a lube oil magnetic chip detector. On the other hand, fractures thru a gear rim (as opposed to thru a gear tooth) are almost always catastrophic and tend to propagate fairly rapidly. This is one reason rotorcraft drivetrain gears use an L2 reliability rate for tooth bending fatigue, while rolling element bearings can safely use an L10 reliability rate for contact fatigue.
One thing I noticed looking at
this cross section of the gearbox is how thin the rim section of the output stage planet gear is. This type of planet gear and bearing configuration having a thin rim without adequate radial stiffness can present a fatigue issue. The planet gear is subject to fairly large opposing inward radial forces twice per rev, from the mesh contacts at the ring gear and sun gear. Basically this means the planet gear rim gets squashed into a slightly oval shape twice per revolution. In my opinion, based simply on what I see in the drawing, it seems like the planet gear rim could have been made a bit thicker. Since this gear is a fracture critical component, I'd think this particular condition of crack propagation from a race surface spall would be a case considered in the fracture analysis.
There is one other thing that seems a bit strange to me based on what I read. If the planet gear rim fracture initiated at a spall in the race surface, there should have been enough ferrous debris generated by the early spall formation to be picked up by a chip detector long before it became a problem. Once again, looking at the cross section drawing of the gearbox linked above, there is an oil tray below the planetary stages. I wonder if the shape of this tray, combined with swirling motion imparted to the return oil flow by windage from the close proximity rotating components, effectively trapped the ferrous debris above the tray and prevented it from reaching the chip detector in the sump.
I'd just like to state the comments above are all just speculation on my part, and I admit to having no detailed knowledge of the design of the gearbox in question.
