We've only had 5 minutes to talk about it so far, but unless something changes, I don't see why we wouldn't go back to M5.

Well having a double server failure would definitely make me think about going back to the same solution..

I'd say the opposite: having experienced this, they would make sure to significantly reduce the probability of it happening again.

In the past, I had a similar problem because of using hardware from the same batch. In retrospect, it's silly to be surprised they died at the same time.