Long before that main rotor shaft suffered structural failure, the MRGB HUMS should have given an indication of a problem.
If only! As I understand it the defect originated in a planet gear which of course as well as rotating itself, is also travelling around the epicyclic. Therefore one point of the gear (where the crack was developing) follows a complex path which only repeats at relatively infrequent intervals. If you understand how the signal averaging techniques (used to extract data on a specific item using external sensors) works, you would know that this problem would result in very long acquisition times to get enough data for the signal averaging to work. That means lots of memory and a requirement to maintain stable flight conditions for a long time. This is not impossible, just difficult and thus not designed into HUMS technology whose concepts were developed in the 1990s.
One of the self-inflicted injuries of HUMS is that, whilst it is good, it is not a panacea and when an accident occurs that HUMS did to detect, it is ridiculed by the disbelievers. All the accidents that would have happened without HUMS are not taken into account.