Disclaimers: I haven't flown a fling-wing in decades, and when I did it was for the Big Green US team, all we had back then was a chips light, driven by several magnetic sensors in the fluid path. My experience with vibration acoustics is also more than 25 years old, so take that as a caution.
Back when I worked with Solar Turbines and Spectral Dynamics, it was relatively easy to monitor, discover, and trigger on FFT characteristics in a turbine and gearbox. I'm considering the rapid detection and prevention of the catastrophic failure which was found in this accident.
By use of the FFT algorithms in the detector hardware and software, we were able to build a characteristic signature of every turbine and gearbox built and instrumented. In some ways, the signatures are very common, and in other ways they are always slightly different. Think of the concept of a fingerprint and one will get the general idea. In the video above, at the 8:15 time mark you can see a brief image of the vibration signature of some rotating asm(I doubt it's a helicopter gearbox, but hard to say). By use of threading and setting up triggers on both amplitude, frequency, and harmonics I'm inclined to say that discovery and alert of an impending failure once spall 3 or spall 4(almost surely) would be possible.
The amount of vibration noise would be of greatest concern, where specific anomalies could present a trigger warning, which would induce a 'false positive' alarm for potential catastrophic failure. The S/N ratio, and type of detection would require sensors inside the MGB, as I think outer case sensors would not be sufficient to capture the ringing and vibration in an encapsulated gearbox bathed in lubricant. This would require further study.
As I mentioned, my depth of knowledge on this is quite old, but back in the 90s we were able to detect common gear spalling in planetary gearboxes for industrial and non-flight aviation systems, most commonly GPU and remote start mechanisms, as well as fixed generators.
Sensor location: At Spectral Dynamics I was a small part of a team which took great interest in instrumentation for failure discovery. There was clear evidence that the location and orientation of the acoustic/vibration sensors for various gearboxes was critical to gathering useful transient data for the electronics to evaluate. For example, using the main mast as the natural Y axis, and the plane of the planetary/sun gears as the X axis, it would seem that having the piezio-electric sensor aligned with the X axis is most effective. However, there are cases where the sensor alignment has shown better response when the vector moment is some degrees off axis, and aligned slightly to the Y axis(something to do with the materials di-pole moment of inertia? Beyond the scope here).
We only got a small taste of what additional sensors, real time analysis is taking place. Which is sad, because I think that even with this type of very small spall propagation it would be possible to detect using internal sensors in the gearbox, and characterization of the vibration signatures in real time. Of course, each gearbox changes its 'tune' with age and strain, but anomalies are really not very hard to detect, and the spalling taking place in what I've seen in the report, could quite possible have been detected, and a warning issued well before the complete failure. I will respectfully say that I have no idea how long spall 1 proceeded to spalls 2, 3, and 4 but my recall of the instances we had at Solar were that it was many operation hours before the system would self-destruct. Also, before anyone points out, I will point out that our sensors, and all gathering and analysis data was ground mounted, and presents a significant weight in a helicopter where weight is the enemy.
YMMV