Pablo - if memory serves, there was a pressure valve in the MRGB line that opened when the normal gearbox pressure dropped - this allowed the feed from the torquemeter pumps (Tq meters worked on oil pressure) to be diverted to both the high speed inputs and the top of the MRGB. No electrical pump required.
Sasless, the FAA accepted SAC's assertion that, since all the oil pathways were inside the MRGB on the S92, there couldn't be a catastrophic loss of oil - that got them a pass/bye/wild card so they didn't have to meet the 30 mins test. Surely someone with such responsibility as the FAA engineers must have asked the question about the filter housing since that was clearly the only external element and should have been subject to some failure mode analysis. To the cynical, it seems like a fudge by a US federal department to help a large US company get their product into the market.